Graph
DTI.
- PYTORCH
- PYG
- RDKIT
- FASTAPI
- SAGEMAKER
Hi there, I’m
Ali Marhon
Building from biology.
Data scientist building ML pipelines for translational oncology and cardiovascular genomics. Geometric deep learning for drug discovery, reproducible single-cell analyses, and production RNA-seq stacks running on Compute Canada HPC.
[ 02 / ABOUT ]
About.
Data scientist working at the intersection of machine learning and biology. My focus is on turning raw biological signal — single-cell transcriptomes, molecular graphs, sequence data — into reproducible, statistically defensible insight: pipelines that re-run end-to-end from raw archives, models that hold up under held-out evaluation, and analyses that respect the hierarchy of the experiment (patient, not cell).
Recent work spans graph deep learning for drug–target interaction prediction and sex-stratified differential expression in cardiovascular disease. The throughline: small, well-instrumented experiments whose conclusions you can actually defend.
Alongside the biology work, I also do quantitative trading research — modeling intraday order flow and limit-order-book dynamics on tick data, training sequence models for short-horizon alpha signals, and stress-testing microstructure-aware execution strategies in walk-forward backtests.
[ 03 / RESEARCH & EXPERIENCE ]
Research
& experience.
-
Mar 2025 — Mar 2026 · Current
Data Scientist / Bioinformatics Analyst
Lawson Health Research Institute · Western University affiliate · London, ON
Building end-to-end RNA-seq and single-cell ML pipelines (STAR, Salmon, DESeq2, Scanpy) for translational oncology — processing 1,200+ patient samples through a Docker-containerized, Snakemake/Airflow-orchestrated, MLflow-tracked stack on Compute Canada HPC.
Recent work includes a random-forest oncology classifier (transcriptomic + clinical, n=480) reaching AUROC 0.87 / PR-AUC 0.71 on a held-out cohort, and a LIMS-to-warehouse ETL (Python + REST + PostgreSQL + dbt) connecting STARLIMS to internal analytics that eliminated ~12 hours/week of manual extraction for the pathology team. Plus gene-set enrichment pipelines (Fisher's exact, BH-FDR, GSEA) and internal R/Python workshops for wet-lab scientists.
-
M.Sc. Biochemistry · Robarts Research Institute
Age- & Sex-dependent Profiling of Angiotensin II–Induced Thoracic Aortic Degeneration in Mice
Pickering Lab · Robarts Research Institute · Western University
Thesis: Age- and Sex-dependent Profiling of Angiotensin II–Induced Thoracic Aortic Degeneration in Mice. Investigated sex- and age-dependent protective mechanisms in murine TAA through integrated wet-lab and computational workflows — in situ aortic measurements, histology, RNA isolation, and transcriptomic profiling on the NanoString nCounter platform.
Built reproducible R pipelines for differential expression, pathway enrichment, and multi-group transcriptomic integration. Identified metabolic and inflammatory programs associated with vascular protection — including altered branched-chain amino acid catabolism and vascular smooth muscle cell remodeling.
-
2019 — 2021
Undergraduate Research Assistant
Western University · London, ON
Functional analysis of ATAD5 — a gene central to DNA replication, repair, and genome maintenance with implications for cancer biology. Generated wild-type and mutant constructs via PIPE cloning, performed site-directed mutagenesis and Sanger sequencing for downstream functional assays.
Day-to-day molecular biology stack: plasmid preparation, bacterial transformation, colony PCR for insertion verification, midi/miniprep extractions, and maintenance of E. coli cultures + competent cells for cloning workflows.
-
Sept 2019 — Apr 2020
Molecular Dynamics of Keap1 R470H — a Cancer-Linked Mutation
Choy Lab · Department of Biochemistry · Western University
Computational structural biology project investigating how the Keap1 R470H mutation — the most prevalent Keap1 alteration in cancer, located distally from the canonical Nrf2 binding pocket — alters protein stability and contributes to constitutive Nrf2 activation and downstream chemoresistance.
Ran all-atom MD simulations in GROMACS with the CHARMM36 forcefield (PDB 2FLU, R470H modeled via in silico mutagenesis, missing atoms reconstructed with MODELLER). Computed radius of gyration, backbone RMSD, per-residue RMSF, and Ramachandran φ/ψ trajectories at residue 470 across 1 μs simulations at 310 K and 410 K — the mutant tracks wild-type at physiological temperature but expands and explores additional conformational basins under thermal stress, consistent with altered allosteric coupling.
-
2018 — 2020
Clinical Research Assistant · Cardiovascular
University Hospital · London Health Sciences Centre · London, ON
Supported a minimally-invasive cardiovascular outcomes study — patient data collection and database maintenance in close collaboration with physicians, nurses, and PIs to ensure accurate documentation and regulatory compliance. Generated anonymized visual reports for internal review and external presentations; handled patient recruitment and follow-up with strict confidentiality.
Worked daily with Cerner and Kroll pharmacy management systems alongside clinical pharmacists on medication documentation, workflow optimization, and EMR integration.
[ 04 / PROJECTS ]
Selected
projects.
01 / 05
← → or click to navigate
[ 05 / CONTACT ]
Let’s
talk.
Open to research collaborations, ML / DS opportunities, and conversations about data, biology, or markets. Email is fastest.