Oncocyrix Multicohort

Oncocyrix Multicohort is a production-grade, Scanpy-based single-cell RNA-seq analysis pipeline designed for 10x Genomics data.

It supports single-sample and multi-cohort studies, with optional CAR-T–aware analysis, cell-type annotation, pseudobulk DESeq2, and multi-database pathway enrichment.

The pipeline is built for research reproducibility, structured outputs, and large-scale cohort integration, including pre/post, multi-patient, and multi-condition study designs.

---

Overview

This repository provides an end-to-end single-cell RNA-seq workflow that combines:

• Standard scRNA-seq best practices (QC, clustering, DE)
• Cohort-aware integration and comparison
• Bulk-grade statistics via pseudobulk aggregation
• Immuno-oncology–focused CAR-T state modeling

It is suitable for both exploratory single-cell analysis and translational / clinical research pipelines.

---

Key Features

Core scRNA-seq Analysis

• 10x Genomics loader (raw mtx/tsv/gz)
• Robust QC & filtering (mitochondrial %, gene counts)
• Normalization, log1p, HVG selection
• Dimensionality reduction (PCA, UMAP, optional t-SNE)
• Clustering & trajectory inference (Leiden, DPT)

Multi-Cohort & Integration

• Single-sample and multi-cohort modes
• Batch correction & integration (BBKNN)
• Group-wise comparisons (e.g., Pre vs Post, Tumor vs Normal)

Cell-Type Annotation

• Metadata-driven annotation (if provided)
• ML-based annotation via CellTypist (optional)
• Cell-type–specific and cluster-specific marker discovery

Differential Expression

• Single-cell DE (Scanpy, Wilcoxon)
• Group-wise DE within clusters or cell types
• Pseudobulk aggregation
• DESeq2 via rpy2 for bulk-grade inference

Pathway Enrichment

• Multi-database enrichment:
  • GO BP / MF / CC
  • KEGG
  • Reactome
  • WikiPathways
• Publication-ready plots
• Semantic pathway deduplication (MiniLM + FAISS, optional)

---

CAR-T Analysis (Optional)

When enabled, the pipeline becomes CAR-T aware, providing biologically interpretable functional state modeling.

Enabling CAR-T

CAR-T analysis is controlled in code, not via CLI:

DO_CART_SCORING = True

Supported CAR-T States

• TStemCM_like
• TPEX_like
• TEX_terminal
• Effector_TEFF
• Proliferating_T
• Terminal_diff

CAR-T Outputs

• CAR-T gene signature scoring per cell
• Refined CAR-T state classification
• CAR-T UMAPs and score visualizations
• Patient × phase × state × gene summaries
• Pre/Post delta tables
• CAR-T state marker genes

---

use pip install oncocyrix_multicohort --mode multi --multi-base-dir "folder location where 10x multi samples are avb"

Installation

Core installation

pip install .

Full installation (recommended)

pip install ".[all]"

R dependencies (for DESeq2)

Requires R ≥ 4.0 with:

install.packages(c("DESeq2", "ggplot2", "pheatmap"))

---

Package Structure

oncocyrix-multicohort/
├── pyproject.toml
├── README.md
└── oncocyrix_multicohort/
    ├── __init__.py
    └── pipeline.py

CLI entry point

oncocyrix-multicohort → oncocyrix_multicohort.pipeline:main

---

Usage

Single-sample mode

oncocyrix-multicohort \
  --mode single \
  --single-10x-dir /path/to/10x/sample \
  --out-name SC_ANALYSIS_RESULTS

Multi-cohort mode

oncocyrix-multicohort \
  --mode multi \
  --multi-base-dir /path/to/GSE208653_RAW \
  --out-name SC_ANALYSIS_RESULTS

See cli.md for full CLI usage and metadata requirements.

---

Output Structure

SC_ANALYSIS_RESULTS/
├── 00_analysis_summary
├── 01_qc_and_filtering
├── 02_highly_variable_genes
├── 03_dimensionality_reduction_and_embeddings
├── 04_clustering_and_cell_states
├── 05_celltype_analysis
├── 05_CART_analysis
├── 06_groupwise_deg
├── 07_pathway_enrichment
├── 08_pseudobulk
├── 09_reference_summary
└── *.h5ad

---

Requirements

• Python ≥ 3.9 (tested up to 3.12)
• R ≥ 4.0 (for DESeq2)
• Recommended RAM ≥ 32 GB

---

Citation

Malik S.
Oncocyrix Multicohort: A CAR-T–aware single-cell RNA-seq analysis framework.

---

Author

Sheryar Malik
Bioinformatics Scientist

---

License

MIT License