ppromprasert/crc-ancestry-transcriptomics-immunomics
Domain:
healthcareCreator:
ppr
Host:
CRC bulk transcriptomics and tumor-immunology analysis integrating African ancestry, CMS1–4, DESeq2, xCell, CIBERSORT, and continuous ancestry gradients.
# CRC Ancestry, Transcriptomics, and Tumor Immunology
**Bulk RNA-seq analysis of African-ancestry-associated transcriptional and immune/stromal variation in colorectal cancer, integrating data-driven ancestry grouping, CMS1–CMS4 stratified DESeq2, continuous ancestry gradients, xCell, CIBERSORT LM22, and immune-marker validation.**
## Scientific question
Does African-ancestry proportion associate with tumor gene-expression and immune/stromal states in colorectal cancer, and are those associations preserved when the analysis is stratified by Consensus Molecular Subtype (CMS)?
## Why this is a separate project from ancestry estimation
The companion population-genomics workflow estimates ancestry.
This repository asks the downstream biological question:
```text
genetic ancestry
|
continuous AFR proportion
|
data-driven AFR grouping
|
bulk tumor RNA-seq
|
+-----+-------------------+
| |
gene expression tumor microenvironment
| |
DESeq2 xCell + CIBERSORT
| |
CMS1 / CMS2 / immune + stromal states
CMS3 / CMS4 |
| |
+-----------+-------------+
|
ancestry-associated
tumor biology
```
## 1. AFR group derivation
The source workflow evaluates **five strategies** for converting continuous AFR ancestry into Low/Medium/High groups:
1. equal tertiles;
2. outer quartiles;
3. Jenks natural breaks;
4. kernel-density minima;
5. mean ± 0.5 SD.
The extended grouping script then applies the **Jenks natural-break solution** and logs the exact cutoffs and provenance.
That is preferable to choosing thresholds solely because they produce visually convenient groups.
Continuous AFR is retained for complementary correlation analyses because any categorization loses information.
## 2. CMS-stratified transcriptomics
The source project contains dedicated ancestry differential-expression scripts for multiple CMS subtypes and uses a common DESeq2 …