Mohammad Salma, PhD

Research Engineer · Bioinformatics & Data Science — 📍 Paris / Montpellier, France

About Me

Research engineer and data scientist with 7 years building bioinformatics pipelines and machine learning systems for genomics, hematopoiesis, and cancer biology.

I design and optimize NGS workflows for single-cell data (scRNA-Seq, scATAC-Seq), bulk sequencing (RNA-Seq, ATAC-Seq, ChIP-Seq, Exome-Seq), long-read sequencing (Oxford Nanopore), and Spatial Transcriptomics — deployed on HPC clusters and cloud environments.

• 🏥 Currently: Project Manager · Institut Imagine, Université Paris Cité
• 🔬 Previously: Senior Bioinformatics & Data Scientist · IRCM Montpellier
• 🧬 Previously: Bioinformatics & Data Scientist · IGMM / CNRS (Eric Soler group)
• 📊 19+ publications · 236 citations · h-index 7

Tech Stack

Programming

Machine Learning & Deep Learning

Workflows & Infrastructure

Apps & Visualization

Featured Projects

RegFiT — Salma M.*, Bouvier Q.* et al. — manuscript in preparation

Machine learning framework for enhancer-based gene regulatory network (eGRN) reconstruction from paired ATAC-seq and RNA-seq data — works with as few as two experimental conditions. RegFiT exploits TF footprinting via TOBIAS to quantify actual TF occupancy at nucleotide resolution, clusters co-varying TFs using WGCNA Dynamic Tree Cut, and applies the RuleFit algorithm to identify sparse, interpretable combinations of TF binding dynamics predictive of differential gene expression. Applied to a temporally resolved erythroid differentiation dataset (8 stages), RegFiT uncovered a novel functional interaction between the canonical erythroid TF KLF1 and the EMT factor ZEB1 — independently validated by genetic perturbation in murine erythroleukemia cells.

First author & co-corresponding author — in collaboration with IGMM/CNRS (Eric Soler group) and LIRMM/CNRS

Var|Decrypt — Salma M. et al., Epigenetics & Chromatin, 2023

Web platform for WES variant exploration, clustering, enrichment analysis, and patient stratification. Built for bench scientists with no bioinformatics background.

ONTdeCIPHER — Cherif E. et al., Bioinformatics, 2022

Nanopore long-read pipeline for real-time pathogenic variant tracking, integrating short-read and long-read data.

GATv2 Multi-Omics Classifier — manuscript in preparation

Graph Attention Network (PyTorch Geometric) for gene regulation classification from TFBS-derived multi-omics graphs, with attention-based interpretability at node and edge level.

NGS Multi-Omics Suite

Production pipelines for RNA-Seq, scRNA-Seq, scATAC-Seq, ChIP-Seq, Exome-Seq, and Spatial Transcriptomics — with automated QC, deployed on HPC and cloud.

Selected Publications

📄 Full list on ORCID · mohammadsalma.com