Single-cell integration in python with scanpy
HTML-код
- Опубликовано: 4 окт 2024
- Simple python single-cell dataset integration using a reference sample. This is an easy integration method that is suitable if you have similar samples. I also show how to compare the proportion of cells between samples and graph it in a simple example.
Notebook:
github.com/mou...
Hi. I'm Ph.D student from South Korea. RUclips algorithm lead me to this precious lectures. Your lectures are so much powerful and helpful! Thank you so much!!
Thank you! Getting comments like this gives me motivation to make more!
I beats me how this channel doesn't have more subscribers and views.
Concise yet clear explainations. Keep up the geat work Mr.Sanbomics! Cheers!
It's getting there! It's been growing steadily for the last 4 or so months. Thank you!
Your channel is great, worth 1 M subs
Thank you! Maybe one day xD
Well explained; thanks for all the great valuable videos.
Thank you for watching them!
I love you bro, you solved a lot of my problems. Thank you very much.
Glad I could help
I’m really looking forward to the video of the advanced integration using scanpy and Scvi; when are you going to upload that?
I plan to do that in the next couple of weeks!
@@sanbomics cool 👍
Uploaded an intro to scVI. Will go over other scVI tools in coming weeks. Hope you find it useful!
@@sanbomics That's great, Thanks for letting me know 👍👍👍
Is the ingest method suitable if there is not a reference sample? And how to identify a reference sample?
If you have multiple samples you can pick one as the reference. Only use this method if the cells are from the same experiment and you expect them to have the same subpopulations. If you expect one might have an extra population, use that as the reference.
thanks for sharing this video! that helps a lot! so at last you mentioned that I need the raw data when you do differential expression. I have no idea about it. should I add raw data before the preprocess?
Depending on what software you are using for DE. If you are just finding cluster markers you don't need to necessarily. You can add the raw to the adata before preprocessing or you can map the annotations back onto the freshly imported data
thanks alot for this tutorial, you cant imagine how much its helpful for us to have this explainations.
i have a question, if i want to intergrate developmental samples of diffrent time laps,( emrbyo of 11 days, 12 days, 13 days,)
which intergration method u recommend, cause i dont know making one "' the youngest" as reference and map others on it
Don't use this method for integrating those samples. Try the integration method I use in either my complete single-cell introduction or in my SCVI introduction (both are the same).
nice video, where can I found these codes so I can just copy and paste ?
Hi! The notebook should be in the video description. Let me know if you have any issues
Hi there, thank you for all your bioinformatics videos, they are super helpful for my research :) I am currently working on annotating using sc.tl.ingest, and I was wondering if it would be possible to inherrit cell-type annotations from observations of a reference AnnData object to a query AnnData instead of just leiden or louvain clusters? For context, the leiden clusters in my reference dataset don't contain the annotations, they are stored in a different observation key and, as far as I know, don't correlate to the clusters detected by any clustering algorithm.
You can map predictions to individual cells with scvi/scarches. ruclips.net/video/tgk-rT_R4wk/видео.html
nice tutorial. i get a basic question, I do not have a 10X_h5 file. I only get 8 matrix files and I need to make an integration analysis. How to do it? Thanks
Hi. Scanpy can open multiple formats. I recommend looking at all the functions starting with scanpy.read_
If it is just a csv you can open it with scanpy.read_csv. But you may have to use a .T to transpose it
can you make a tutorial about scanorama? , I need to know why we use the highly variable genes in the integration instead of all genes of my samples, if i used the highly variable genes i will lose my rest genes. can you explain this issue please
Hi. Possibly in the future, I haven't used scanorama. You should use the raw data that includes all the genes for differential expression. Integration is best for annotation and visualization. It is still an ongoing challenge in the field and can bias gene expression etc.
Hello! Could you please tell how did you animation for your logo?
1) Chunk the photo. 2) Turn chunk into linear array. 3) Run UMAP on the chunks where each chunk is a sample. 4) Find before/after for each chunk, i.e., where was it in the picture and where did it end up in the UMAP. 5) Plot a line from each before and after position. 6) Move a dot which is the average color of the chunk along the line 100 times to make 100 unique figures. 7) Put the figures together in a GIF/video. For the logo part "Sanbomics" I did the same thing but I created a mask for the letters in a picture and randomly assigned an after spot to the points which fell somewhere on the mask. I threw it together in a script if you want it (not polished and you have to get dependencies).
@@sanbomics Awesome! Thank you for your response. it would be super helpful to have a script, I need it for my 5-min PhD pitch :)