Code repository accompanying the preprint:
Robotic perturbation proteomics and AI agents enable scalable drug mechanism discovery
Yuming Jiang, Cameron S. Movassaghi, Jesús Muñoz-Estrada, Niveda Sundararaman, Amanda Momenzadeh, Jesse G. Meyer
bioRxiv (2026)
Preprint: https://www.biorxiv.org/content/10.64898/2026.05.04.722718v1
This repository contains analysis notebooks, scripts, AI-generated outputs, manual-check results, clinical validation analysis, figure-generation code, and web-portal code associated with a large-scale HepG2 drug perturbation proteomics study.
This project develops an end-to-end platform that combines semi-automated sample preparation, rapid LC-MS/MS proteomics, and AI-agent-assisted data analysis for scalable drug mechanism discovery.
In the associated study, 172 compounds were profiled in HepG2 cells with six biological replicates per compound, generating 1,232 proteomes and quantifying 8,703 proteins.
AI results/ AI-generated analysis outputs ClinicalValidation/ EHR validation analyses Figures/ Figure-generation scripts MultiAgent/ GPT-based multi-agent interpretation workflow data/ Example and processed data manual check/ Human-verified outputs website/ Streamlit portal
- Proteomics data processing and quality control
- Protein filtering
- Sample filtering
- Missing-value assessment
- Missing-value imputation
- Batch correction
- QC visualization
- Differential protein regulation analysis
- Welch's t-test
- Benjamini-Hochberg FDR correction
- Significant proteins: FDR < 0.05 |log2FC| > 0.5
- Pathway and mechanism-of-action analysis
- Hallmark pathways
- GO Biological Process pathways
- GSEA analysis
- Drug clustering
- Mechanism inference
- AI-agent-assisted interpretation
- Molecular identity profiles
- Drug-level summaries
- Pathway interpretation
- Known MoA recovery
- Novel hypothesis generation
- Validation recommendations
- Multi-agent framework
- Analyzer
- Reviewer
- Supervisor
-
Clinical validation Includes EHR-based validation workflows.
-
Interactive website https://drug-ai-robotics-proteomics.streamlit.app/
MassIVE accession: MSV000101671
FTP: ftp://massive-ftp.ucsd.edu/v12/MSV000101671/
Interactive website: https://drug-ai-robotics-proteomics.streamlit.app/
Python 3.13.9
Common dependencies: numpy pandas scipy statsmodels scikit-learn matplotlib seaborn gseapy networkx streamlit
BASIC USAGE
git clone https://github.com/xomicsdatascience/AI-robotics-proteomics.git
Some code and outputs were generated or assisted by AI systems. All major analytical logic and outputs were manually reviewed.
Jiang Y, Movassaghi CS, Muñoz-Estrada J, Sundararaman N, Momenzadeh A, Meyer JG.
Robotic perturbation proteomics and AI agents enable scalable drug mechanism discovery
bioRxiv 2026.
DOI: 10.64898/2026.05.04.722718
Yuming Jiang yuming.jiang@csmc.edu
Jesse G. Meyer jesse.meyer@csmc.edu