@@That_Chemist Really interesting. Seems like the falsified NMR data at least could have been caught. Editors absolutely need to be on the lookout for cover ups, hand drawn baselines AND copy/pasting sections of spectra (especially from spectra within the same paper lol).
Yes, all this add up to the noise that makes studies more expensive and the results more uncertain. If you cannot be honest, then science is not for you.
As a non-chemist (with computer stuff as a hobby), I was really surprised by how clearly visible the falsification was once you know what to look for. Like, I'm sure given a few hours, a basic understanding of how NMR files are formatted, and some code, you could make much more plausible looking noise to fill in what you want to hide, and it would surprise me if someone hasn't done just that.
@@RyanTosh yes, but remember: who is falsifying data? its not some guy that goes above and beyond expextations to do something good but just someone dishonest who just wants to crank out that paper. Scientists arent magicians or superhumans. Theres bad apples among them
This kind of stuff arises directly from the academic mentality of publishing "discoveries" at all costs. Journals of negative results really need to be more prominent in all areas! Hiding failed experiments only slows scientific progress.
Plus if someone already tried your incorrect idea and it didn’t work, there’s a massive value in not duplicating the work, especially as a new researcher or grad student wasting your time.
I know this is a chemistry channel, but it honestly baffles me some of the falsified western blots that get published in pharmacology/molecular biology/biochemistry journals. Sometimes researchers will literally go into windows paint (or something similar) and draw black lines on the blot to represent the protein bands! I don't understand how reviewers don't immediately spot this. Love this series anyhow. Hope to see more in the toxic series soon!
Falsifying scientific data is just as bad if not worse than pretending to be differently abled(i.e. disabled) for clout. Just because you worked on something for long time, but didn't get correct data don't mean that you should falsify the data. Just publish that what you thought was wrong. every single bit of information helps. If in future, someone wants to work on something you did, and while reading through the literature they find out its not possible in a certain way, they can come up with ways that actually work, instead of basing their research on false things
>Falsifying scientific data is just as bad if not worse than pretending to be differently abled(i.e. disabled) for clout. Falsification of scientific date is clearly much, much, much, much worse. It should be a criminal offense. Science is already mostly BS nowadays, even outside of sociology. Research is valuable, but high standards must be maintained. Feels weird calling not falsified data "high standard".
@@vornamenachname594 It might have good sides though: it can encourage more people for verifying stuff on their own and do their own research instead of blindly believing what the priests of science say. That's the primary reason why I started watching this channel.
I think it would be better to solve this problem by removing the main reason for publishing false data. With "publish or perish" there is an immense pressure on everyone besides those lucky few with a permanent employment contract resulting in desperate mistakes like falsifying data.
@@user255 Not really. No real journal will publish your collection of failed attempts synthesizing compound xy without a very good reason. Especially if you do something like total synthesis you can run in the problem that at the end of your funding you dont have anything publishable. If you dont think that the pressure of having to publish is the reason what reason would you see in publishing falsified data? Inpatients? Normally it is just a question of time to get some nice data. What reason do you see that people are falsifying data besides time pressure?
@@Seorful Like I said, negative studies are under appreciated. If the methods you used in trying to synthesize some compound were plausible/sensible, then it would be really helpful for others to learn what doesn't work in order to avoid doing the same non-working things again and again. This would also take off the pressure, because then you wouldn't have to succeed in order to publish. And I'm not talking only about chemistry.
@@user255 I'm lacking experience to talk about fields outside of chemistry. Also I dont think that it will be easier to get journals to accepts negative studies than to just pay researchers a fair amount and give them a permanent employment contract so that they can focus on actual research without having to worry if they will be without a job should they not finish their research fast enough.
@@Seorful Example in medicine negative studies are already accepted and trials are forced to be registered (this is requirement of some journals), so that negative results wont be lost in file drawer. But I think we also need what you suggest.
What's funny is I tell my friend that not all papers are absolutely true, yet he solely believes that all research papers are 100% correct and factual... this is a very good example, glad I found your channel! Keep up the good work!
One other hilarious/sad thing I came across, was the SI of a paper about biomass conversion. The authors had great conversions/selectivities to useful products, yet the paper was in a low impact journal. Upon looking at their GC data I realised why that was. The authors just turned off the FID detector for a solid 5 minutes within their GC method, so the undesired products didn't show up. Then the detector was conveniently on again when their desired products came over. Of course you can easily tell, since the baseline was at 0.00 for those solid 5 minutes whereas typically the baseline was 10-20
Oh wow, I actually worked on this project (after it became clear what had happened and the findings needed to be reproduced), along with some continuations. Figured it would've come up eventually since you're a fluorochemistry guy. Small world!
Well, the main reason for falsification of spectra is the pointless race after super clean text book quality spectra. If there is some residual toluene in a compund, which will be then reacted in toluene, what is bad about that? Also if there are are some inseparable imputies, it is better to honestly show them rather then eploy some other unspecified purification techniques. I trust papers that have dirty NMRs.
For practical purposes it doesn’t matter, but for analytical purposes it is always best to remove solvent if you are able to - if removal of solvent isn’t possible without decomposition, then the presence of solvent is justified
You are doing a huge community service by showing examples of bad data and bad papers. All undergrads need to have several lectures in their field like this so they know how to spot field specific issues.
Really enjoying this series. It is sadly even more common in life sciences/biology. Not to close any doors this early in my career, but I don't see myself ever touching cancer because literally 90% of results can't be replicated, and that includes from nature, science, cell etc. Hot fields like cancer get staggering amounts of money (I know a lab that got a $3,000,000 direct cost RO1 and the NIH came back and asked them to write a supplement so they could justify giving them another $400,000). Meanwhile in rare disease land, we have the chance to provide a treatment for pediatric diseases that have 0 options and have to fight tooth and nail for grants. The pressure is very high, thankfully not from my lab or boss, or university but I know it exists elsewhere. I can only imagine the pressure is 10x worse at a small school without a big name like mine, or 100x worse at a school outside us/aus/Western Europe/China/japan
I am the only remaining student in the Weaver Lab who remembers Mo. He was the first one to demonstrate his reaction set ups for me, a first-year grad student in early 2019.
Unfortunately, there are much more sofisticated methods to alter or erase some peaks in raw NMR. These are often unnoticed by community, because in SI, there are always pictures which cannot be zoomed that much you can see altered part. I don´t know why it isn´t mandatory to create one .zip file with all raw .fid files which can be opened anytime. Aside of checking if the spectra are legit, it has added benefit of actually seeing fine structure and coupling constants of complicated multiplets etc.
From a school stoichometry quiz "What is a mole: A. A mammal known for digging up gardens. B. A small dark spot on the skin. C. A spy embedded in an enemy government. D. 6.02*10^23 particles of a substance. E. All of the above."
Fully Agree with you re holding publishers to account on this. They have the entire industry bent over a barrel but cant afford to carry out proper QC checks? Behave
You should be able to trust that the author of any publication has enough scientific integrity to present the data in a truthful way. There are so many ways that people can cheat, and a lot of them are difficult to catch as a reviewer. Using NMR as an example: the latest versions of mestrenova have a function to hide peaks and if you turn your spectrum into a picture (even a high resolution one) it is impossible to see. The only way to make sure would be to require the submission of the original data, and even then it can be that the fid was manipulated
There should be digital signature for the raw data. And the same thing for cameras and raw pictures. We already use digital signatures, why not to apply them more broadly!
@@jimcrelm9478 Of course not. Firstly, private keys are just like any private data (family photos etc). 100% compatible with all GNU and general open source licenses. And in fact all major Linux distros support and use them already. No need for TPM. If you would tamper the firmware you would also destroy the encrypted keys. Secondly, I don't think current NMR devices run open source firmware..? I wish they would. Thirdly, I never suggested that this would replace requirement for replication. That would be insane.
@@jimcrelm9478 Of course the machine does the signing. Why that would need proprietary binary? If the device runs Linux, then simple bash script run as root could do it. Fully open source. I don't follow your thinking at all.
@@jimcrelm9478 Of course not. The script is run as root and the key is owned by root. Everything in encrypted file system. You would operate user space of the device. You are arguing against standard things in Linux. PS. I don't think you understand what digital signing means. Look up how it is done in PGP.
I think this could be solved through some kind of authentication of the spectra data. Something like embedding a public key in the NMR machine to encrypt the data and send it to some kind of online service provided by the machine manufacturer that generates a spectra "link" that can be viewed. Something similar to a digital signature
@@user255 you need a way to guarantee that the user can't tamper with the data, either locally or in-transit and a way to validate the signature. This is usually done through public/private key-pairs extensively in software/hardware industry to sign data. So essentially you would send the signed content/file that is cryptographed by the NMR machine that has the public key stored safely in hardware (to avoid you signing a fake spectra locally and sending that to verification) to this manufacturer platform that has the private key and can authenticate/verify the data sent and generate a link that can be used to view the spectra by the publications. But I am not sure that I understood your question and what you are considering a "digital signature"
@@igorghignatti Look up example how PGP signature works. You can check the authenticity locally, if you have the public key. The NMR device could use key provided by the manufacturer.
Whats stoping mister dishonest researcher from hooking up to the sensor wires and feeding bad data to the machine?Then they can go "look it is signed,it must be good,stop acusing me!"
A closely-related issue that's rarely been talked about has to do with the general policy on SIs. Most journals don't have to review them (although plenty of referees do) prior to publication. I've seen tons of SIs that might not contain false data, but they look like they were made by a bunch of high school kids. Badly organised, inconsistent formatting, etc. In my opinion, the SI is as important (if not more important) as the main text. It's in the name, it literally supports the main paper, so it should be taken more seriously by both authors and publishers!
You can do a type 3 fake but much better. He honestly did a very poor job here. But even submitted high quality data could be forged. Ideally you'd have the NMR machine digitally sign & stamp the data, which you could still bypass, but it's a very different skillset than fudging graphs.
In my opinion, chemists should attached the MestreNova format file in their SI world file. So the PI or reviewers could click the spectrum like ChemDraw pic to check
@@liverpoolirish208 yeah, no proprietary industry standards pls. FID files would be best for verifcation and if i recall correctly you can just drag and drop them into mnova so its still accessible to those that just use that
I can't share many details, but I have had collaborators share baseline and live-phase bleed samples with us for analysis, but not the necropsy samples, because the data didn't look as good in the pilot study. Dishonest as well, in my opinion (their molecule didn't work anyway).
This practice just worsen the problem in current research. More and more papers aims to "fill a yearly quota" becuase of economic pressures in the research system. The ratio between reviewers and submitted papers is decreasing. Finally, it just worsen the spread of misinformation info to the public, even good and well checked specific idea or proposal in a researches sometimes turns into "generalized fact totally based on this article from this research, please buy this snake oil " its like a game of telephone.
My first assignment at UTSA was to recover spectral data from a dead hard drive on a GE 300 MHz NMR. The Geniuses at Univ had had indications oh impending hard drive failure and only THEN mattempted to make a backup of the drive OVER an older, good backup copy. The project ended with many months of spectra irretrievably lost. Emphasize to your students that when they have data, it is Their responsibility to make backup copies of the data, both processed and raw. Do not count on IT people to keep things running smoothly. Do NOT count on them to preserve your data. A pretty common method for IT to handle a troublesome computer it yo re-image the computer, thus destroying any data it once held. Your data is YOUR responsibility, and no one else's.
Hi, I'm studying machine learning and I'd just like to recommend that journals not use machine learning to detect falsified data, or at the very least have everything looked over by a human in addition. Machine learning systems can be very smart at times, but they can often be VERY stupid, and often in ways that not even a human would be. In this case I'd be worried about false negatives more than false positives, as I'd assume everything marked as possible violations would be reviewed by a human. Machine learning is not something that can or should replace human analysis.
Falsifying scientific data can lead to bad laws that hurt the economy in the most extreme example. These spectra I believe are computer files and computer files have associated metadata. Some metadata is contained by the file and some is associated with the filesystem. The students should make sure they keep this in tact. Having the drive where the file was originally created is you best protection. Pushing files up to the cloud may lose or change the metadata. It should be backed their as there is always the possibility of the original drive being destroyed corrupted or stolen. In the recent Depp-Heard case missing metadata caused Heard some grief. According to Attorney Christopher Melcher you don't through your old phones out as they contain evidence which you may need down the road. I think care needs to be taken of the spectral data files that become part of an publication.
As an ex-chemist (ABD but with a masters) and current data scientist, I'm super hesitant about "create a machine learning algorithm to automatically analyze spectra" if there is no human validation. Machine Learning is cool, but I find that people who aren't involved in the ML space SUBSTANTIALLY overestimate how accurate they are, and underestimate how easy it is to adversarially feed them bad data when it benefits you, etc. In my chem days I wasn't too involved in spectroscopy, so I don't have "published paper" level of experimental knowledge, but I do have at least "entry level PhD lab class" level experience with doing some syntheses and then doing basic UV-Vis or Raman spec, and knowing how complex they can get when you're trying to suss out impurities, contamination, etc... I think it's too high-dimensional of a question to really entirely trust to an algorithm. You could definitely train an ML Algo to filter out some of the real shitty cases, but ehh...
05:55 How about using a logarithmic scale or something of that sort? That should crank up the tiny parts of the spectrum without messing up the tall peaks, right? 08:08 Then someone will probably create a neural network for generating spectra that "look plausible" for the publisher's bots :) (and probably for humans as well).
i dont know anything about chemistry (i only had like 3 years in high school), but what i DO know about is programming. I know for someone outside the immediate field its enticing to think "just do some machine learning on that and we solved it", but unfortunately you first need data, lots of it, from many sources. The data needs to actually be analyzable (eg. raw files, no jpeg, png or anything - you COULD do it with images, but then you can ONLY do images), you need to verify that what the model is training is actually the right thing, not just some superficial data that arises from, eg. when the file was made, or with which specific NMR apperatus (i dont know the lingo too well, so forgive me). then your training data needs to be a good sample (proportions need to match), which is where it gets REALLY tricky: you need falsified data, that occurred naturally (you cant just go and make some yourself, the machine will only learn how to spot your fakes). and even after all of that, you still need to just have that system "flag" something as suspicious for review by a human - this is to spot possible false positives (would be a great way to ruin your journal's reputation afterall). and for continued operation you still need new samples. sorry for the long comment. tldr: Machine Learning is harder then people think, at least doing it properly
I'm in the weaver group. I just had a few things to add here haha... 1) great job of handling this professionally without throwing the whole group under the bus. Our group is bustling and some really exciting work is being done here. 2) During our weekly meetings, he meticulously points out any inconsistencies in our spectra now. I don't know how it was when Mo was here (as I was not here at the time), but now he's rigorous, and it's still a sore subject for him that we comedically bring up even still. 3) Mo was a post-doc which makes it tem times worse, it should have never happened. Also, Mo got fired from his industry job after this happened. 4) Jimmie was the one that discovered and told JACS, if not, it may still be published to this day. 5) it was all done in microsoft paint lmao. We were SO embarrassed, and honestly why didn't the editors catch this? Jimmie is a very busy man, and he does a better job about this now, but it's literally the editors job to check things like this. Cheers
Also to anyone justifying false data (e.g., "my PI expects too much from me" etc.), There is no excuse and it will never pay off in the long run. It is a sign of bad character and any excuse around that is moot. Falsifying data should never cross a successful chemist's mind.
Great video as always @That_Chemist ! However is strongly disagree with your take that the corresponding author is not responsible for what happened. He is the corresponding author, so it his obligation to verify the data that his lab publishes. I am a PI myself and i check everything before something associated with my name gets out. People make mistakes and that is normal, its the PIs duty to train people how to do good science and also make sure no fraud happens. This whole thing has been a massive display of oversight by the PI in my opinion.
Publishers take advantage of creators to make money. The key tool they use to do this is they maintain a lack of accountability. They will never spend time criticizing submissions themselves.
I was working with dibromoxylenes... Made by xylene NBS bromation and washing the reaction mixture with water... Product was crystalized from hexanes:ethanol mixture and CDCl3 NMR showed tall, slim peak aroud 3 ppm (1H NMR). We all thought it was a water (even my supervisor). But next reaction with those was giving like super low yields. Expected about 70 %, found 30-ish. We were clueless... Found out a few months later, synthetising those nasty bromated xylenes again, that its not water... It was succinimide that wanst really washed out as paper said. Funny was that it crystalized out with our product. It was lowering the yields because we were doing secondary amine + xylene... Sooo yeah. Even when there is an easy answer to a peak, try other possible outcomes. If it was just wet (because ethanol used for crystalization was just standard denaturated, not dry one), yields would've been much higher.
Never before have I heard a RUclips channel plug a researcher’s papers before. XD Excellent work! You’re doing us all a service, the greatest glory of science is that it can be proven…. Or it’s inverse. Looking forwards to all of your content!!!
Oh boy, the first few spectra were ridiculous. How did any of those pass a review process? Even the last one was really low effort, just zeroing part of the data. Dishonesty like this should really be punished harshly. For fields where modelling of what your signal is going to look like is necessary to analyze your data, generating false datasets which are indistinguishable in a plot and often in raw data as well, is super easy. So it is necessary for there to be trust in the scientific community and we shouldn't be easy on people who want to sell fake data.
I don't know how useful it would be for other people but I think it would be interesting to see what a research chemist's workflow looks like. I'm guessing that the cycle from "not having an idea for what you want to do" all the way to having a paper published is very different than it would be in, say, computer science or mathematics or condensed matter physics. I'm sure its very different depending on what kind of chemistry you're doing and who you're working for but I don't have even a general idea of how it works. This might get TL;DR from here: As an example of what I'm talking about, I can give an example from computer science. One day I was wondering why spiders don't appear to have compound eyes like a lot of other arthropods, and after doing some research, they don't have compound eyes. Specifically things with large eyes like jumping spiders and huntsman have one lens and multiple retinas at different focal distances. The spider's brain integrates this information to produce depth perception with only one eye necessary. That was enough for me to say "how can I implement this in a machine vision context". Next came simple static shots (i.e. not video) and test inputs where I knew what the output should be and writing simple programs to give and verify that output. After that, I decided to use two cameras with coaxial optical paths and a beam splitter to give different focal lengths. The code went from Perl to C with x86 assembly and after several rounds of optimization and bug fixing I had a finished product. I didn't write a paper but I suspect I could have. Some of the impetus for this came from having worked on stereoscopic machine vision in the past. If you made it this far, I guess my point is this: chemistry is way harder than computer science. I know a lot of chemistry now, but I don't have the 25 years of experience I do with computer science to know what my work cycle should be, even as an amateur. I'm kind of past the point of copying other peoples' syntheses and I'm getting past the point of just doing derivatives of those syntheses (i.e. playing pin the alcohol on the salicylate and seeing what it smells like). I'd rather not have it take another 20 years for me to get to where I want to be with chemistry, so even just like a link to "this is what you should do" would be helpful. I'm sorry if this question amounts to "tell me all of chemistry", but thats why I gave the example in the middle. I hope the general question makes sense.
That would be a good video idea. Speaking personally as a pretty inexperienced researcher, it's always good to, like in any context, start from a question. A pretty common one is to look at a pre-existing synthesis and say something like "Why is it that all of these steps are necessary?, *are* all of these steps necessary? What kind of synthetic steps would need to exist to make a better synthesis here?" For instance, in the paper mentioned in this video (which I worked on a little bit after the retraction), the goal was to develop a synthetic step that could condense a lot of syntheses that require specific substitution patterns on fluoro-arenes. So you might say, "Okay, can I reliably use this method to make this pattern?, How about that pattern?" Then go through and see which ones succeed and fail. A great example of this is shown in That Chemist's video (I think it's on this channel too, but it's for sure on his old one) where he goes through one of his PhD projects and does a similar thing for thionoesters.
@@SaberToothPortilla I'm just having a hard time understanding what a university research chemist does. I can understand commercial research chemistry a bit better and pharmaceutical research better than that (I was raised by two research biologists, so there's no getting away from that). I'm looking at resinferatoxin, which is, to me, too complex to comprehend and I guess the syntheses is really quite difficult. (it is to capsaicin what a nuclear bomb is to a hand grenade). I think there's a paper on its synthesis and maybe that would help with looking for ways to simplify things, but my gut reaction is to say "make a computer figure it out". I'll have to think about it some more. I haven't watched his video on his PhD work yet because I want to watch a few other things first. I also don't really understand intentionally focusing on sulfur and fluorine compounds. I know its not, but it seems arbitrary, like saying "I'm only going to work with VW Bug hubcaps and Hello Kitty refrigerator magnets from now on". Maybe its just a mindset thats just incompatible with my own. Sorry if this comment came off kind of random but uh... that's probably because it kind of is.
@@htomerif Well, it's similar to how you might get particularly well versed in certain coding paradigms or machine learning methods. Others are useful and valuable, and you know what they do, but you pick your domain of focus/work and you get good at it. An academic research chemist, broadly speaking, will do similar things to what commercial and pharmaceutical research chemists will do, depending on their domain of focus of course. Of course, in a commercial context there tends to be more focus placed on discovery (you can't really sell it if it isn't new), but you can patent syntheses too, so even in that context it comes up. In academics there's usually more of a focus on, well, the chemistry. Finding novel chemical processes that can used to improve things or, occasionally, produce things that couldn't be produced before. Nano materials have been getting more and more prominent. Thin film application and other fabrication relevant stuff. Environmental Chemistry. It's a broad field, and there's plenty that isn't *just* focused on synthesis. It's just that Organic Chemisty research usually comes back to synthesis in particular. Also, yeah, the synthesis of resiniferatoxin is evidently quite involved. Apparently 25+ significant steps. Even just looking at it, it's a much more complicated compound than capsaicin for sure. For what it's worth, the synthesis was developed in 1997, but it was also done by a Stanford group, so, certainly not simple.
@@SaberToothPortilla I watched That Chemist's recent video on failing and that gave me a much better understanding of what the work cycle is, at least for him. Its basically 1, notice some similarities between some things. 2, try to adapt an existing process to a new set of chemicals. 3, shoot every conceivable iteration of that process at the problem and see if you win (and how much you win). 4, fail or publish and go back to 1. So really there's a lot of intuition involved in every step, and in my primitive way thats what I've been doing by tacking alcohols onto salicylate and seeing what they smell like. Another example was my coming across citalopram (an SSRI) somewhere and noticing its similarity to doxylamine and diphenhydramine. No science on my part but just reading about it, it turns out my intuition about citalopram being an antihistamine was correct. Its really hit or miss though, for me.
In the lab of the late Dr Leroy Townsend NMR data was necessary but not sufficient to prove a structure of a compound. UV data was also obtained, when needed, which could answer questions of the substitution pattern of a particular nitrogen heterocycle. For compounds that were to be submitted for biological testing I required for my students and myself to get a clean H-NMR before submitting for elemental analysis ( which were cumulatively expensive.). If the H-NMR was close to maximum load and the spectrum was clean the compound was submitted for EA. If it passed then it was submitted for testing. If not the compound was repurified. For us purity was paramount. Mass Specs gives molecular weights but does it really determine purity? Your video provokes the thought that methods papers should be treated like a submission to Org Syn and independent verification by a second lab should be done in place of peer review.
I honestly think it should be, especially if the authors aren't going to film and discuss the results of their experiments for at least a portion of them.
Well as researchers pay publishers to publish and readers pay them for access, money shouldn't be the issue. And especially peer-review etc is done pro-bono anyway.... So publishers probably just cut corners with the paif editing.
I am getting grease impurities in my NMR despite distillation of PET ether. Are impurities from grease acceptable in JOC or OL? What to do to remove such impurities? Where am I going wrong? P. S: Even after washing with pentane, grease impurities remains.
Let me get this straight. Academic journals don't require you to submit raw experimental data? Kids at my high school science fair were disqualified if they didn't provide the data used for a single graph. How the fuck is a high school science fair more diligent than an academic journal? O.o
I’m still an undergrad so my experience is obviously limited, but personally in most organic papers I’ve seen, spectra usually aren’t provided at all. Most of the time all I see is a list of all the peaks ie δ(1H, 600MHz): 4.35 (1H, dd), 1.24 (3, s) etc. Does it depend on the journals whether the actual spectra are required?
I’ve worked under Dr. Weaver, however I was working on other research project at the time. He is an excellent guy and I’m honestly surprised that he didn’t notice this, but oversights do happen. No matter the circumstances though, these actions should be unacceptable.
ah, science. the replication crisis marks the corrupt minds that are now all over the scientific fields im happy that i decided to become a welder rather than head for university was it biology thats 50% replication? shouldnt 50% on its own invalidate all research in that particular field by the scientific method- which relies exactly on replication?
I picked out some bad data from a practice lab that 100% resulted from a bug in the software. My boyfriend who's in math was like... "No you can't do that" oh you sweet summer child lol.
How checked out do you have to be to just white box over lmao.Falsifing this kind of data by writing a python script to generate a dirty baseline with whatever peaks you want seems trivial.
You should look into the work of Dr Elizabeth Bik, if you haven't already. She does a lot of image forensics of biological blots, and scientific dishonesty .
ahh, I feel like this is corny af but hear me out xP ..Im practicing my note-taking, right.. [I have issues keeping shit concise :S] and I've been doing this with a few vids here n there..this being one of them.. so that brings me to printing the transcript so I can continue my note-taking later on but also take it a bit slower..and well, I was thinking of how much better it would be if I could just print ur slides :)) while I'm aware that you walk us through everything, I won't necessarily have your narration later :P so..you know.. the outline helps tie things together a bit better than just the transcript -- so...just wanted to say: would be cool if there were an option to print ur slides! I suppose I can just print a screenshot [of each slide], but anyway-.... food for thought, maybe? :]] ^__^
I have the actual ppt slides for my main lecture series available as a patreon perk, but if people wanted the other slides I could add those too - it’s a bit of a higher price because it’s meant for professionals, but if you decide to pursue it you should message me on discord/patreon :)
You explained that Jimmie Waver has nothing to do with the problem or the other authors. I would have liked you to do the same with Prof Cossy I know her personally and worked with her and I know the true story and what happened it is also the responsibility of one of the author and he is the same on all the mentioned papers. it would help if you were more scientific and objective.
I would still say that every author on any paper has the responsibility to review the data in any paper that they are an author on, but I didn’t emphasize any responsibility on Janine specifically. The difference with Jimmie is that I heard him openly discuss it prior to giving a research presentation, and he was forthright and open about it. I have not had the pleasure of hearing Janine speak on it, so I will withhold further comment, although it sounds like you are fairly passionate about it, and I think the scientific community would benefit from people such as yourself making content surrounding topics they are passionate about
@@That_Chemist If you say so, then why you are excluding Jimmie and his co-authors from responsibility? For me also all authors should be responsible. we shall stand on principles or we shall not stand at all. No thanks, you seemed passionately defending Jimmie in your video and kept on insisting during the second part that is only the leading author's responsibility.
You say to not blame the corresponding author, because he is a nice person?? Being a corresponding author should mean something. Remember ''with a great power comes great responsibility'' :).
@@That_Chemist Is that really enough? When you look to other instances in your video, the authors take a collective responsibility and do not blame just 1 person. I think that is more fair. The supervisor should be able to control the people working under him/her, and never allow this type of things to happen. Please do not get me wrong, you did a great job by highlighting this nasty topic.
Maybe you should also put the example of most recent retracted Science paper (Iridium-Catalyzed Asymmetric Borylation).
A link to the article in question www.science.org/doi/10.1126/science.abq4856
@@That_Chemist Really interesting. Seems like the falsified NMR data at least could have been caught. Editors absolutely need to be on the lookout for cover ups, hand drawn baselines AND copy/pasting sections of spectra (especially from spectra within the same paper lol).
The fact that these falsifications are so blatant makes me wonder how many falsified papers exist that are undetectable
Thia practice is so cash money
Yes, all this add up to the noise that makes studies more expensive and the results more uncertain. If you cannot be honest, then science is not for you.
As a non-chemist (with computer stuff as a hobby), I was really surprised by how clearly visible the falsification was once you know what to look for. Like, I'm sure given a few hours, a basic understanding of how NMR files are formatted, and some code, you could make much more plausible looking noise to fill in what you want to hide, and it would surprise me if someone hasn't done just that.
@@RyanTosh yes, but remember: who is falsifying data?
its not some guy that goes above and beyond expextations to do something good but just someone dishonest who just wants to crank out that paper. Scientists arent magicians or superhumans. Theres bad apples among them
I recommend you to check out the plastic fantastic scandal because that’s crazy
I'll say it once again, but this series is of absolute public interest. Thank you for giving some people tools to fight scientific dishonesty
Thanks :)!
This kind of stuff arises directly from the academic mentality of publishing "discoveries" at all costs. Journals of negative results really need to be more prominent in all areas! Hiding failed experiments only slows scientific progress.
Plus if someone already tried your incorrect idea and it didn’t work, there’s a massive value in not duplicating the work, especially as a new researcher or grad student wasting your time.
I know this is a chemistry channel, but it honestly baffles me some of the falsified western blots that get published in pharmacology/molecular biology/biochemistry journals. Sometimes researchers will literally go into windows paint (or something similar) and draw black lines on the blot to represent the protein bands! I don't understand how reviewers don't immediately spot this. Love this series anyhow. Hope to see more in the toxic series soon!
you should make videos covering them!
Or like the WBs where the authors blatantly just duplicated the same image and mirrored it or sth
Falsifying scientific data is just as bad if not worse than pretending to be differently abled(i.e. disabled) for clout. Just because you worked on something for long time, but didn't get correct data don't mean that you should falsify the data. Just publish that what you thought was wrong. every single bit of information helps. If in future, someone wants to work on something you did, and while reading through the literature they find out its not possible in a certain way, they can come up with ways that actually work, instead of basing their research on false things
That was the case when science was about truth, not money grants and fame.
@@bonbonpony yep. it is the need for grant money drives a lot of bad science imo
>Falsifying scientific data is just as bad if not worse than pretending to be differently abled(i.e. disabled) for clout.
Falsification of scientific date is clearly much, much, much, much worse. It should be a criminal offense. Science is already mostly BS nowadays, even outside of sociology. Research is valuable, but high standards must be maintained. Feels weird calling not falsified data "high standard".
@@vornamenachname594 It might have good sides though: it can encourage more people for verifying stuff on their own and do their own research instead of blindly believing what the priests of science say. That's the primary reason why I started watching this channel.
@@vornamenachname594 Oh, and BTW, RUclips ate your comment.
I think it would be better to solve this problem by removing the main reason for publishing false data. With "publish or perish" there is an immense pressure on everyone besides those lucky few with a permanent employment contract resulting in desperate mistakes like falsifying data.
I don't think that is the reason. If you do research you have always something to publish. The problem is that negative studies are under appreciated.
@@user255 Not really. No real journal will publish your collection of failed attempts synthesizing compound xy without a very good reason. Especially if you do something like total synthesis you can run in the problem that at the end of your funding you dont have anything publishable. If you dont think that the pressure of having to publish is the reason what reason would you see in publishing falsified data? Inpatients? Normally it is just a question of time to get some nice data. What reason do you see that people are falsifying data besides time pressure?
@@Seorful Like I said, negative studies are under appreciated. If the methods you used in trying to synthesize some compound were plausible/sensible, then it would be really helpful for others to learn what doesn't work in order to avoid doing the same non-working things again and again. This would also take off the pressure, because then you wouldn't have to succeed in order to publish.
And I'm not talking only about chemistry.
@@user255 I'm lacking experience to talk about fields outside of chemistry. Also I dont think that it will be easier to get journals to accepts negative studies than to just pay researchers a fair amount and give them a permanent employment contract so that they can focus on actual research without having to worry if they will be without a job should they not finish their research fast enough.
@@Seorful Example in medicine negative studies are already accepted and trials are forced to be registered (this is requirement of some journals), so that negative results wont be lost in file drawer. But I think we also need what you suggest.
What's funny is I tell my friend that not all papers are absolutely true, yet he solely believes that all research papers are 100% correct and factual... this is a very good example, glad I found your channel! Keep up the good work!
Great point!
I think that Voltaire would say that your friend "lives in the best of all possible worlds." The bliss of naivety is overpowering.
One other hilarious/sad thing I came across, was the SI of a paper about biomass conversion. The authors had great conversions/selectivities to useful products, yet the paper was in a low impact journal. Upon looking at their GC data I realised why that was. The authors just turned off the FID detector for a solid 5 minutes within their GC method, so the undesired products didn't show up. Then the detector was conveniently on again when their desired products came over. Of course you can easily tell, since the baseline was at 0.00 for those solid 5 minutes whereas typically the baseline was 10-20
Send me a link to the article!
'Publish or perish' is the culprit, every time.
Oh wow, I actually worked on this project (after it became clear what had happened and the findings needed to be reproduced), along with some continuations.
Figured it would've come up eventually since you're a fluorochemistry guy. Small world!
cool! small world!!
This is my favorite series hands down
Thanks for letting me know :D
Well, the main reason for falsification of spectra is the pointless race after super clean text book quality spectra. If there is some residual toluene in a compund, which will be then reacted in toluene, what is bad about that? Also if there are are some inseparable imputies, it is better to honestly show them rather then eploy some other unspecified purification techniques. I trust papers that have dirty NMRs.
For practical purposes it doesn’t matter, but for analytical purposes it is always best to remove solvent if you are able to - if removal of solvent isn’t possible without decomposition, then the presence of solvent is justified
You are doing a huge community service by showing examples of bad data and bad papers. All undergrads need to have several lectures in their field like this so they know how to spot field specific issues.
Thank you!
NMR: my profs always wanted the NMR and the integration on top of the spectra.
A GC/MS is hard to beat for structure confirmation.
I used GC-MS a ton to confirm structures, it is definitely underutilized
Really enjoying this series. It is sadly even more common in life sciences/biology. Not to close any doors this early in my career, but I don't see myself ever touching cancer because literally 90% of results can't be replicated, and that includes from nature, science, cell etc. Hot fields like cancer get staggering amounts of money (I know a lab that got a $3,000,000 direct cost RO1 and the NIH came back and asked them to write a supplement so they could justify giving them another $400,000). Meanwhile in rare disease land, we have the chance to provide a treatment for pediatric diseases that have 0 options and have to fight tooth and nail for grants. The pressure is very high, thankfully not from my lab or boss, or university but I know it exists elsewhere. I can only imagine the pressure is 10x worse at a small school without a big name like mine, or 100x worse at a school outside us/aus/Western Europe/China/japan
:(
I am the only remaining student in the Weaver Lab who remembers Mo. He was the first one to demonstrate his reaction set ups for me, a first-year grad student in early 2019.
Wow - you’re the second one from his lab to comment on this video today!
Unfortunately, there are much more sofisticated methods to alter or erase some peaks in raw NMR. These are often unnoticed by community, because in SI, there are always pictures which cannot be zoomed that much you can see altered part. I don´t know why it isn´t mandatory to create one .zip file with all raw .fid files which can be opened anytime. Aside of checking if the spectra are legit, it has added benefit of actually seeing fine structure and coupling constants of complicated multiplets etc.
From a school stoichometry quiz
"What is a mole:
A. A mammal known for digging up gardens.
B. A small dark spot on the skin.
C. A spy embedded in an enemy government.
D. 6.02*10^23 particles of a substance.
E. All of the above."
My favorite high school chemistry teacher gave us extra credit if we stitched together a plushie mole for Mole Day Oct. 23 🤣
Fully Agree with you re holding publishers to account on this. They have the entire industry bent over a barrel but cant afford to carry out proper QC checks? Behave
I’m just an undergrad student, but this does make me wonder about papers that I use as references for essay assignments. (Not a chemistry student FYI)
I’m really surprised. As a biochemist, I’m used to having to provide raw gels, blots, raw numerical values etc. for publication.
You should be able to trust that the author of any publication has enough scientific integrity to present the data in a truthful way. There are so many ways that people can cheat, and a lot of them are difficult to catch as a reviewer. Using NMR as an example: the latest versions of mestrenova have a function to hide peaks and if you turn your spectrum into a picture (even a high resolution one) it is impossible to see. The only way to make sure would be to require the submission of the original data, and even then it can be that the fid was manipulated
yeah :(
There should be digital signature for the raw data. And the same thing for cameras and raw pictures. We already use digital signatures, why not to apply them more broadly!
@@jimcrelm9478 Of course not.
Firstly, private keys are just like any private data (family photos etc). 100% compatible with all GNU and general open source licenses. And in fact all major Linux distros support and use them already. No need for TPM. If you would tamper the firmware you would also destroy the encrypted keys.
Secondly, I don't think current NMR devices run open source firmware..? I wish they would.
Thirdly, I never suggested that this would replace requirement for replication. That would be insane.
@@jimcrelm9478 Of course the machine does the signing. Why that would need proprietary binary? If the device runs Linux, then simple bash script run as root could do it. Fully open source. I don't follow your thinking at all.
@@jimcrelm9478 Of course not. The script is run as root and the key is owned by root. Everything in encrypted file system. You would operate user space of the device. You are arguing against standard things in Linux.
PS. I don't think you understand what digital signing means. Look up how it is done in PGP.
I think this could be solved through some kind of authentication of the spectra data. Something like embedding a public key in the NMR machine to encrypt the data and send it to some kind of online service provided by the machine manufacturer that generates a spectra "link" that can be viewed. Something similar to a digital signature
Excellent idea !!
Why not just digital signature?
@@user255 you need a way to guarantee that the user can't tamper with the data, either locally or in-transit and a way to validate the signature. This is usually done through public/private key-pairs extensively in software/hardware industry to sign data. So essentially you would send the signed content/file that is cryptographed by the NMR machine that has the public key stored safely in hardware (to avoid you signing a fake spectra locally and sending that to verification) to this manufacturer platform that has the private key and can authenticate/verify the data sent and generate a link that can be used to view the spectra by the publications. But I am not sure that I understood your question and what you are considering a "digital signature"
@@igorghignatti Look up example how PGP signature works. You can check the authenticity locally, if you have the public key. The NMR device could use key provided by the manufacturer.
Whats stoping mister dishonest researcher from hooking up to the sensor wires and feeding bad data to the machine?Then they can go "look it is signed,it must be good,stop acusing me!"
A closely-related issue that's rarely been talked about has to do with the general policy on SIs. Most journals don't have to review them (although plenty of referees do) prior to publication.
I've seen tons of SIs that might not contain false data, but they look like they were made by a bunch of high school kids. Badly organised, inconsistent formatting, etc. In my opinion, the SI is as important (if not more important) as the main text. It's in the name, it literally supports the main paper, so it should be taken more seriously by both authors and publishers!
I totally agree - I wish every substrate always had its own full write up, rather than just saying "it was prepared according to General Procedure 1"
As someone with little to no experience with these graphs, I could probably pick up how to spot fakes pretty quickly.
A fun read, especially if you aren't familiar with BS-NMR - www.reddit.com/r/chemistry/comments/sk7i96/alternative_strategies_for_nmr_processing_and/
You can do a type 3 fake but much better. He honestly did a very poor job here. But even submitted high quality data could be forged. Ideally you'd have the NMR machine digitally sign & stamp the data, which you could still bypass, but it's a very different skillset than fudging graphs.
holy shit how could these people not see how obvious their own fake was
It amazes me
In my opinion, chemists should attached the MestreNova format file in their SI world file. So the PI or reviewers could click the spectrum like ChemDraw pic to check
The file would heavily lagg on some computers
Mestrenova is expensive, proprietary software. Perhaps the raw FID would be better.
@@liverpoolirish208 yeah, no proprietary industry standards pls. FID files would be best for verifcation and if i recall correctly you can just drag and drop them into mnova so its still accessible to those that just use that
I can't share many details, but I have had collaborators share baseline and live-phase bleed samples with us for analysis, but not the necropsy samples, because the data didn't look as good in the pilot study. Dishonest as well, in my opinion (their molecule didn't work anyway).
:(
If they cannot be honest, then science is not for them. You should blow the whistle.
I just love that angry Benzene, is so aestetik
Credit to JPepper
Just wanna say I appreciate that you provide links for the DOI's
Thanks!
This practice just worsen the problem in current research. More and more papers aims to "fill a yearly quota" becuase of economic pressures in the research system. The ratio between reviewers and submitted papers is decreasing. Finally, it just worsen the spread of misinformation info to the public, even good and well checked specific idea or proposal in a researches sometimes turns into "generalized fact totally based on this article from this research, please buy this snake oil " its like a game of telephone.
My first assignment at UTSA was to recover spectral data from a dead hard drive on a GE 300 MHz NMR. The Geniuses at Univ had had indications oh impending hard drive failure and only THEN mattempted to make a backup of the drive OVER an older, good backup copy. The project ended with many months of spectra irretrievably lost. Emphasize to your students that when they have data, it is Their responsibility to make backup copies of the data, both processed and raw. Do not count on IT people to keep things running smoothly. Do NOT count on them to preserve your data. A pretty common method for IT to handle a troublesome computer it yo re-image the computer, thus destroying any data it once held. Your data is YOUR responsibility, and no one else's.
Hi, I'm studying machine learning and I'd just like to recommend that journals not use machine learning to detect falsified data, or at the very least have everything looked over by a human in addition. Machine learning systems can be very smart at times, but they can often be VERY stupid, and often in ways that not even a human would be. In this case I'd be worried about false negatives more than false positives, as I'd assume everything marked as possible violations would be reviewed by a human. Machine learning is not something that can or should replace human analysis.
I think we need both
I think you overestimate the amount of time a person is going to spend on reviewing a paper. Which is a problem.
yes it is
Falsifying scientific data can lead to bad laws that hurt the economy in the most extreme example.
These spectra I believe are computer files and computer files have associated metadata. Some metadata is contained by the file and some is associated with the filesystem. The students should make sure they keep this in tact. Having the drive where the file was originally created is you best protection. Pushing files up to the cloud may lose or change the metadata. It should be backed their as there is always the possibility of the original drive being destroyed corrupted or stolen.
In the recent Depp-Heard case missing metadata caused Heard some grief. According to Attorney Christopher Melcher you don't through your old phones out as they contain evidence which you may need down the road.
I think care needs to be taken of the spectral data files that become part of an publication.
As an ex-chemist (ABD but with a masters) and current data scientist, I'm super hesitant about "create a machine learning algorithm to automatically analyze spectra" if there is no human validation. Machine Learning is cool, but I find that people who aren't involved in the ML space SUBSTANTIALLY overestimate how accurate they are, and underestimate how easy it is to adversarially feed them bad data when it benefits you, etc. In my chem days I wasn't too involved in spectroscopy, so I don't have "published paper" level of experimental knowledge, but I do have at least "entry level PhD lab class" level experience with doing some syntheses and then doing basic UV-Vis or Raman spec, and knowing how complex they can get when you're trying to suss out impurities, contamination, etc... I think it's too high-dimensional of a question to really entirely trust to an algorithm.
You could definitely train an ML Algo to filter out some of the real shitty cases, but ehh...
yeah, at least if it goes "hey this one looks suspicious editor, take a look at it"
As a chemistry freshman, I love this channel!
05:55 How about using a logarithmic scale or something of that sort? That should crank up the tiny parts of the spectrum without messing up the tall peaks, right?
08:08 Then someone will probably create a neural network for generating spectra that "look plausible" for the publisher's bots :) (and probably for humans as well).
i dont know anything about chemistry (i only had like 3 years in high school), but what i DO know about is programming.
I know for someone outside the immediate field its enticing to think "just do some machine learning on that and we solved it", but unfortunately you first need data, lots of it, from many sources.
The data needs to actually be analyzable (eg. raw files, no jpeg, png or anything - you COULD do it with images, but then you can ONLY do images), you need to verify that what the model is training is actually the right thing, not just some superficial data that arises from, eg. when the file was made, or with which specific NMR apperatus (i dont know the lingo too well, so forgive me).
then your training data needs to be a good sample (proportions need to match), which is where it gets REALLY tricky: you need falsified data, that occurred naturally (you cant just go and make some yourself, the machine will only learn how to spot your fakes).
and even after all of that, you still need to just have that system "flag" something as suspicious for review by a human - this is to spot possible false positives (would be a great way to ruin your journal's reputation afterall). and for continued operation you still need new samples.
sorry for the long comment.
tldr: Machine Learning is harder then people think, at least doing it properly
Yeah the problem is people don’t have sufficient literacy in multiple domains
Unless you’re bill gates, and then you know basically everything
I'm in the weaver group. I just had a few things to add here haha...
1) great job of handling this professionally without throwing the whole group under the bus. Our group is bustling and some really exciting work is being done here.
2) During our weekly meetings, he meticulously points out any inconsistencies in our spectra now. I don't know how it was when Mo was here (as I was not here at the time), but now he's rigorous, and it's still a sore subject for him that we comedically bring up even still.
3) Mo was a post-doc which makes it tem times worse, it should have never happened. Also, Mo got fired from his industry job after this happened.
4) Jimmie was the one that discovered and told JACS, if not, it may still be published to this day.
5) it was all done in microsoft paint lmao. We were SO embarrassed, and honestly why didn't the editors catch this? Jimmie is a very busy man, and he does a better job about this now, but it's literally the editors job to check things like this.
Cheers
Also to anyone justifying false data (e.g., "my PI expects too much from me" etc.), There is no excuse and it will never pay off in the long run. It is a sign of bad character and any excuse around that is moot. Falsifying data should never cross a successful chemist's mind.
I totally agree - you’re the third person from his lab to comment today!
Thanks a lot. Can't wait to use this guide to falsify data on my next paper.
(jk)
NOOO 🤣🤣🤣
thank you, this will be very useful for when I need to falsify NMR data
Great video as always @That_Chemist ! However is strongly disagree with your take that the corresponding author is not responsible for what happened. He is the corresponding author, so it his obligation to verify the data that his lab publishes. I am a PI myself and i check everything before something associated with my name gets out. People make mistakes and that is normal, its the PIs duty to train people how to do good science and also make sure no fraud happens. This whole thing has been a massive display of oversight by the PI in my opinion.
Publishers take advantage of creators to make money. The key tool they use to do this is they maintain a lack of accountability. They will never spend time criticizing submissions themselves.
Great video, keep it coming!!
Thanks! Will do!
I was working with dibromoxylenes... Made by xylene NBS bromation and washing the reaction mixture with water... Product was crystalized from hexanes:ethanol mixture and CDCl3 NMR showed tall, slim peak aroud 3 ppm (1H NMR). We all thought it was a water (even my supervisor). But next reaction with those was giving like super low yields. Expected about 70 %, found 30-ish. We were clueless... Found out a few months later, synthetising those nasty bromated xylenes again, that its not water... It was succinimide that wanst really washed out as paper said. Funny was that it crystalized out with our product. It was lowering the yields because we were doing secondary amine + xylene... Sooo yeah. Even when there is an easy answer to a peak, try other possible outcomes. If it was just wet (because ethanol used for crystalization was just standard denaturated, not dry one), yields would've been much higher.
Never before have I heard a RUclips channel plug a researcher’s papers before. XD
Excellent work! You’re doing us all a service, the greatest glory of science is that it can be proven…. Or it’s inverse.
Looking forwards to all of your content!!!
Thank you!
Oh boy, the first few spectra were ridiculous. How did any of those pass a review process? Even the last one was really low effort, just zeroing part of the data.
Dishonesty like this should really be punished harshly. For fields where modelling of what your signal is going to look like is necessary to analyze your data, generating false datasets which are indistinguishable in a plot and often in raw data as well, is super easy. So it is necessary for there to be trust in the scientific community and we shouldn't be easy on people who want to sell fake data.
I don't know how useful it would be for other people but I think it would be interesting to see what a research chemist's workflow looks like. I'm guessing that the cycle from "not having an idea for what you want to do" all the way to having a paper published is very different than it would be in, say, computer science or mathematics or condensed matter physics. I'm sure its very different depending on what kind of chemistry you're doing and who you're working for but I don't have even a general idea of how it works.
This might get TL;DR from here:
As an example of what I'm talking about, I can give an example from computer science. One day I was wondering why spiders don't appear to have compound eyes like a lot of other arthropods, and after doing some research, they don't have compound eyes. Specifically things with large eyes like jumping spiders and huntsman have one lens and multiple retinas at different focal distances. The spider's brain integrates this information to produce depth perception with only one eye necessary. That was enough for me to say "how can I implement this in a machine vision context". Next came simple static shots (i.e. not video) and test inputs where I knew what the output should be and writing simple programs to give and verify that output. After that, I decided to use two cameras with coaxial optical paths and a beam splitter to give different focal lengths. The code went from Perl to C with x86 assembly and after several rounds of optimization and bug fixing I had a finished product. I didn't write a paper but I suspect I could have. Some of the impetus for this came from having worked on stereoscopic machine vision in the past.
If you made it this far, I guess my point is this: chemistry is way harder than computer science. I know a lot of chemistry now, but I don't have the 25 years of experience I do with computer science to know what my work cycle should be, even as an amateur. I'm kind of past the point of copying other peoples' syntheses and I'm getting past the point of just doing derivatives of those syntheses (i.e. playing pin the alcohol on the salicylate and seeing what it smells like). I'd rather not have it take another 20 years for me to get to where I want to be with chemistry, so even just like a link to "this is what you should do" would be helpful.
I'm sorry if this question amounts to "tell me all of chemistry", but thats why I gave the example in the middle. I hope the general question makes sense.
That would be a good video idea.
Speaking personally as a pretty inexperienced researcher, it's always good to, like in any context, start from a question.
A pretty common one is to look at a pre-existing synthesis and say something like "Why is it that all of these steps are necessary?, *are* all of these steps necessary? What kind of synthetic steps would need to exist to make a better synthesis here?"
For instance, in the paper mentioned in this video (which I worked on a little bit after the retraction), the goal was to develop a synthetic step that could condense a lot of syntheses that require specific substitution patterns on fluoro-arenes.
So you might say, "Okay, can I reliably use this method to make this pattern?, How about that pattern?" Then go through and see which ones succeed and fail.
A great example of this is shown in That Chemist's video (I think it's on this channel too, but it's for sure on his old one) where he goes through one of his PhD projects and does a similar thing for thionoesters.
@@SaberToothPortilla I'm just having a hard time understanding what a university research chemist does. I can understand commercial research chemistry a bit better and pharmaceutical research better than that (I was raised by two research biologists, so there's no getting away from that).
I'm looking at resinferatoxin, which is, to me, too complex to comprehend and I guess the syntheses is really quite difficult. (it is to capsaicin what a nuclear bomb is to a hand grenade). I think there's a paper on its synthesis and maybe that would help with looking for ways to simplify things, but my gut reaction is to say "make a computer figure it out". I'll have to think about it some more.
I haven't watched his video on his PhD work yet because I want to watch a few other things first. I also don't really understand intentionally focusing on sulfur and fluorine compounds. I know its not, but it seems arbitrary, like saying "I'm only going to work with VW Bug hubcaps and Hello Kitty refrigerator magnets from now on".
Maybe its just a mindset thats just incompatible with my own.
Sorry if this comment came off kind of random but uh... that's probably because it kind of is.
@@htomerif Well, it's similar to how you might get particularly well versed in certain coding paradigms or machine learning methods. Others are useful and valuable, and you know what they do, but you pick your domain of focus/work and you get good at it.
An academic research chemist, broadly speaking, will do similar things to what commercial and pharmaceutical research chemists will do, depending on their domain of focus of course.
Of course, in a commercial context there tends to be more focus placed on discovery (you can't really sell it if it isn't new), but you can patent syntheses too, so even in that context it comes up. In academics there's usually more of a focus on, well, the chemistry. Finding novel chemical processes that can used to improve things or, occasionally, produce things that couldn't be produced before. Nano materials have been getting more and more prominent. Thin film application and other fabrication relevant stuff. Environmental Chemistry.
It's a broad field, and there's plenty that isn't *just* focused on synthesis. It's just that Organic Chemisty research usually comes back to synthesis in particular.
Also, yeah, the synthesis of resiniferatoxin is evidently quite involved. Apparently 25+ significant steps. Even just looking at it, it's a much more complicated compound than capsaicin for sure. For what it's worth, the synthesis was developed in 1997, but it was also done by a Stanford group, so, certainly not simple.
@@SaberToothPortilla I watched That Chemist's recent video on failing and that gave me a much better understanding of what the work cycle is, at least for him.
Its basically
1, notice some similarities between some things.
2, try to adapt an existing process to a new set of chemicals.
3, shoot every conceivable iteration of that process at the problem and see if you win (and how much you win).
4, fail or publish and go back to 1.
So really there's a lot of intuition involved in every step, and in my primitive way thats what I've been doing by tacking alcohols onto salicylate and seeing what they smell like.
Another example was my coming across citalopram (an SSRI) somewhere and noticing its similarity to doxylamine and diphenhydramine. No science on my part but just reading about it, it turns out my intuition about citalopram being an antihistamine was correct. Its really hit or miss though, for me.
In the lab of the late Dr Leroy Townsend NMR data was necessary but not sufficient to prove a structure of a compound. UV data was also obtained, when needed, which could answer questions of the substitution pattern of a particular nitrogen heterocycle. For compounds that were to be submitted for biological testing I required for my students and myself to get a clean H-NMR before submitting for elemental analysis ( which were cumulatively expensive.). If the H-NMR was close to maximum load and the spectrum was clean the compound was submitted for EA. If it passed then it was submitted for testing. If not the compound was repurified. For us purity was paramount. Mass Specs gives molecular weights but does it really determine purity?
Your video provokes the thought that methods papers should be treated like a submission to Org Syn and independent verification by a second lab should be done in place of peer review.
I honestly think it should be, especially if the authors aren't going to film and discuss the results of their experiments for at least a portion of them.
I love the soviet quote near the end... made me laugh out loud
Do sad how people forget to cite OG papers in hydrodefluorination. Much worse to fabricate data on a topic very close to my heart
Fighting the good fight.
Awesome video.
Thanks!
Nice video!
Thanks!
Well as researchers pay publishers to publish and readers pay them for access, money shouldn't be the issue. And especially peer-review etc is done pro-bono anyway.... So publishers probably just cut corners with the paif editing.
I am getting grease impurities in my NMR despite distillation of PET ether. Are impurities from grease acceptable in JOC or OL? What to do to remove such impurities? Where am I going wrong?
P. S: Even after washing with pentane, grease impurities remains.
There is probably a contaminating source - probably the rotovap bump trap
@@That_Chemist okay thankyou. Will take care.
Let me get this straight. Academic journals don't require you to submit raw experimental data? Kids at my high school science fair were disqualified if they didn't provide the data used for a single graph. How the fuck is a high school science fair more diligent than an academic journal? O.o
It’s a joke sometimes
I’m still an undergrad so my experience is obviously limited, but personally in most organic papers I’ve seen, spectra usually aren’t provided at all. Most of the time all I see is a list of all the peaks ie δ(1H, 600MHz): 4.35 (1H, dd), 1.24 (3, s) etc. Does it depend on the journals whether the actual spectra are required?
any reputable journal should have spectra always (in the modern age); for older journals, they may not, but that is due to a limitation of technology
I’ve worked under Dr. Weaver, however I was working on other research project at the time. He is an excellent guy and I’m honestly surprised that he didn’t notice this, but oversights do happen. No matter the circumstances though, these actions should be unacceptable.
Watching this video while taking NMRs, lol!
Man these guys are really bad at photoshop. I'd need about 1 min to be a waay better job
ah, science. the replication crisis marks the corrupt minds that are now all over the scientific fields
im happy that i decided to become a welder rather than head for university
was it biology thats 50% replication? shouldnt 50% on its own invalidate all research in that particular field by the scientific method- which relies exactly on replication?
There are lots of issues with replication in science :(
I picked out some bad data from a practice lab that 100% resulted from a bug in the software. My boyfriend who's in math was like... "No you can't do that" oh you sweet summer child lol.
How checked out do you have to be to just white box over lmao.Falsifing this kind of data by writing a python script to generate a dirty baseline with whatever peaks you want seems trivial.
See, stuff like this gives me less confidence in “the scientists”.
and if other scientists put up with it, you should have less confidence in them too! this especially goes for the publishers!!!
You should look into the work of Dr Elizabeth Bik, if you haven't already.
She does a lot of image forensics of biological blots, and scientific dishonesty
.
You should post links to her work in the discord!
It should exist a Journal , " The Journal of Irreproducible results".
ahh, I feel like this is corny af but hear me out xP ..Im practicing my note-taking, right.. [I have issues keeping shit concise :S] and I've been doing this with a few vids here n there..this being one of them.. so that brings me to printing the transcript so I can continue my note-taking later on but also take it a bit slower..and well, I was thinking of how much better it would be if I could just print ur slides :)) while I'm aware that you walk us through everything, I won't necessarily have your narration later :P so..you know.. the outline helps tie things together a bit better than just the transcript -- so...just wanted to say: would be cool if there were an option to print ur slides! I suppose I can just print a screenshot [of each slide], but anyway-.... food for thought, maybe? :]] ^__^
I have the actual ppt slides for my main lecture series available as a patreon perk, but if people wanted the other slides I could add those too - it’s a bit of a higher price because it’s meant for professionals, but if you decide to pursue it you should message me on discord/patreon :)
You explained that Jimmie Waver has nothing to do with the problem or the other authors. I would have liked you to do the same with Prof Cossy I know her personally and worked with her and I know the true story and what happened it is also the responsibility of one of the author and he is the same on all the mentioned papers. it would help if you were more scientific and objective.
I would still say that every author on any paper has the responsibility to review the data in any paper that they are an author on, but I didn’t emphasize any responsibility on Janine specifically. The difference with Jimmie is that I heard him openly discuss it prior to giving a research presentation, and he was forthright and open about it. I have not had the pleasure of hearing Janine speak on it, so I will withhold further comment, although it sounds like you are fairly passionate about it, and I think the scientific community would benefit from people such as yourself making content surrounding topics they are passionate about
@@That_Chemist If you say so, then why you are excluding Jimmie and his co-authors from responsibility? For me also all authors should be responsible. we shall stand on principles or we shall not stand at all. No thanks, you seemed passionately defending Jimmie in your video and kept on insisting during the second part that is only the leading author's responsibility.
You say to not blame the corresponding author, because he is a nice person?? Being a corresponding author should mean something. Remember ''with a great power comes great responsibility'' :).
Which is why when he found out about the issue, he contacted the journal to have the paper retracted!
@@That_Chemist Is that really enough? When you look to other instances in your video, the authors take a collective responsibility and do not blame just 1 person. I think that is more fair. The supervisor should be able to control the people working under him/her, and never allow this type of things to happen. Please do not get me wrong, you did a great job by highlighting this nasty topic.