Webapp#78
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
MikeLippincott
requested review from
gwaybio and
jaceybronte
and removed request for
gwaybio
gwaybio
left a comment
gwaybio
left a comment
There was a problem hiding this comment.
A few, mostly housekeeping comments. @jaceybronte should be the one to give final approval.
Is it possible to make the app run on the github.io pages for this repo (rather than the streamlit.app)?
Also, the app link you sent should be specified in the readme (and it is currently down)
| # - `CRISPRGeneEffect.parquet`: The data in this document are the Gene Effect Scores obtained from CRISPR knockout screens conducted by the Broad Institute. Negative scores notate that cell growth inhibition and/or death occurred following a gene knockout. Information on how these Gene Effect Scores were determined can be found [here](https://genomebiology.biomedcentral.com/articles/10.1186/s13059-021-02540-7) | ||
| # - `depmap_gene_meta.tsv`: Genes that passed QC and were included in the training model for Pan et al. 2022. We use this data to filter genes as input to our models. The genes were filtered based 1) variance, 2) perturbation confidence, and 3) high on target predictions based on high correlation across other guides. | ||
| # | ||
| # > Pan J, Kwon JJ, Talamas JA, Borah AA, Vazquez F, Boehm JS, Tsherniak A, Zitnik M, McFarland JM, Hahn WC. Sparse dictionary learning recovers pleiotropy from human cell fitness screens. Cell Syst. 2022 Apr 20;13(4):286-303.e10. doi: 10.1016/j.cels.2021.12.005. Epub 2022 Jan 31. PMID: 35085500; PMCID: PMC9035054. |
There was a problem hiding this comment.
this is the webster citation and probably should be kept (see line 11)
|
|
||
| # Load depmap metadata | ||
| gene_meta_df = pd.read_parquet(qc_gene_file, sep="\t") | ||
| gene_meta_df = pd.read_csv(qc_gene_file, sep="\t") |
There was a problem hiding this comment.
why would this change to a read_csv?
| @@ -0,0 +1,3 @@ | |||
| version https://git-lfs.github.com/spec/v1 | |||
There was a problem hiding this comment.
i'm concerned with this approach, mostly because i don't understand what will happen when someone uses the webapp. If someone uses the webapp, does it trigger a gitlfs pull? It must retrieve the data from somewhere. Rather than triggering this (which consumes git lfs tokens $$), please use a different approach, of which, there are a few (e.g., figshare)
| @@ -0,0 +1,88 @@ | |||
| altair==6.1.0 | |||
There was a problem hiding this comment.
this dependency list is extremely strict and fragile - almost certainly will break very soon. What is the convention for streamlit? Is it possible to relax restrictions? How is streamlit tested? Consider digging into this a bit more, which will likely increase longevity of the app
| name: isort (python) | ||
| args: ["--profile", "black", "--filter-files"] | ||
|
|
||
| #Code formatter for both python files and jupyter notebooks |
There was a problem hiding this comment.
consider running precommit on this file (and all other files in this repo) as well
| | [0.data-download](0.data-download/) | Download required files | Download gene effect data and cell line information, and download gene QC and construct gene filtering dictionary | | ||
| | [1.data-exploration](1.data-exploration/) | Explore and visualize data | Create figures to visualize cell line information and split gene effect data into balanced test and train dataframes | | ||
| | [2.train-VAE](2.train-VAE/) | Train Beta VAE and Beta TC VAE models | Optimize hyperparameters and train Beta Variational Autoencoder/Beta Total Correlation Variational Autoencoder with optimal hyperparameters and previously created test and train dataframes | | ||
| | [3.analysis](3.analysis/) | Analyze Beta VAE and Beta TC VAE Outputs | Generate heatmaps to visualize death windows by cell line and by genes, run Gene Set Enrichment Analysis with BVAE and BTCVAE synthesized data, and analyze extracted BVAE/BTCVAE latent space data to compare similarity of cancer between different demographics | |
There was a problem hiding this comment.
this comment likely for @jaceybronte - please update the README to mirror what we've now done. This is out of date
jaceybronte
left a comment
jaceybronte
left a comment
There was a problem hiding this comment.
Code looks good, still trying to figure out how to make it work with all this data. LMK if you need me to edit the parquets
| | [0.data-download](0.data-download/) | Download required files | Download gene effect data and cell line information, and download gene QC and construct gene filtering dictionary | | ||
| | [1.data-exploration](1.data-exploration/) | Explore and visualize data | Create figures to visualize cell line information and split gene effect data into balanced test and train dataframes | | ||
| | [2.train-VAE](2.train-VAE/) | Train Beta VAE and Beta TC VAE models | Optimize hyperparameters and train Beta Variational Autoencoder/Beta Total Correlation Variational Autoencoder with optimal hyperparameters and previously created test and train dataframes | | ||
| | [3.analysis](3.analysis/) | Analyze Beta VAE and Beta TC VAE Outputs | Generate heatmaps to visualize death windows by cell line and by genes, run Gene Set Enrichment Analysis with BVAE and BTCVAE synthesized data, and analyze extracted BVAE/BTCVAE latent space data to compare similarity of cancer between different demographics | |
There was a problem hiding this comment.
Are we going to run into conflicts when I merge my PR?
3 participants
This PR is for the webapp in development for Jacey's first author manuscript.