https://github.com/gibsramen/winter-rotation-2019
Knight Lab Rotation Winter 2019
https://github.com/gibsramen/winter-rotation-2019
Last synced: 2 months ago
JSON representation
Knight Lab Rotation Winter 2019
- Host: GitHub
- URL: https://github.com/gibsramen/winter-rotation-2019
- Owner: gibsramen
- Created: 2019-02-08T21:24:42.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2019-03-15T17:18:31.000Z (about 6 years ago)
- Last Synced: 2025-01-19T20:37:41.386Z (4 months ago)
- Language: R
- Size: 163 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Knight Lab Rotation Winter 2019
## Gibraan Rahman
The purpose of this project is to identify relationships between immune cell abundances and newly-discovered microbial communities in TCGA cancers.
### Features:
* Perform Similarity Network Fusion and Consensus Clustering on OTUs/immune cells
* Test cluster results for associations with race, sex, age
* Perform survival analysis on resulting clusters comparing overall survival
* PERMANOVA differential composition analysis of clusters
* (TODO) Differential gene expression using GSEA
### SNF-CC
Create consensus clusters of patients from fused network of microbial profiles and CIBERSORT-inferred immune cell abundances. Performed 9 times with number of clusters = 2:10. Silhouette widths of each of these clusters numbers is saved along with the grouping of each sample.
Input the CIBERSORT file to use and output directory.
Currently trying different pre-processing procedures on the CIBERSORT data to improve clinical relevant of clustered.
Usage:
`qsub -v CIBFILE=,OUTDIR= SNF_CC_all.pbs`
### Covariate Testing
Post-hoc analysis of clustering results to ensure that samples are not simply being separated by race, age, or sex. Inputs are cancer type and number of clusters.
TODO: Create submittable script to perform on group file rather than input cancer type and number of clusters.
Usage:
`Rscript covariate_stats.R `
### Survival Analysis
Create survival analysis plots for all cancer types (all cluster numbers). This analysis uses updated metadata from TCGA (days to death OR right-censored days to last follow-up) to create survival curves.
TODO: Parameterize this procedure such that a group file can be input instead of a cancer type. This will allow for more flexibility in analysis.
Usage:
`qsub survival_analysis.pbs`
### Alpha Diversity Comparison
Compare alpha diversity among patients in each identified cluster.
Usage:
`qsub -v GROUP_FILE= compare_alpha_diversity.pbs`
### PERMANOVA
Perform PERMANOVA on cluster results to analyze differential abundances among clusters.
Usage:
`qsub -v GROUP_FILE= permanova.pbs`
### Feature ANOVA
Perform ANOVA on all features and association with identified group. Right now only supports CIBERSORT features.
TODO: Implement OTU as well.
Usage:
`Rscript feature_anova.R `