Postdoctoral Researcher · DTU BRIGHT, Denmark

I build biology
at a scale you can
actually reason about.

Computational biologist. I built the largest strain-resolved collections of in-silico bacterial cells assembled anywhere, and the pipelines that put that scale within reach of other people. Now I am teaching machines to do the reasoning on top of it.

Explore the work Download CV
0 E. coli strain-specific
genome-scale models
0 Lactobacillaceae
genome-scale models
0 building computational
biology at scale
0 open tools, models
and databases shipped

Act I

From binary to a bioreactor

It starts as bits. Zeros and ones pair off into bases and a genome assembles itself out of information. Everything after this point is rebuilt from public assemblies, so anyone can take the same inputs and get the same answer.

Input
binary stream
Output
1 genome

01 — Work

Models, pipelines and databases

Everything below is public and runnable. The models are browsable in a web viewer, the pipelines are open source, and the databases are citation-backed.

Flagship database First author

BiGG 2026

biggr.org

The next-generation release of the BiGG Models database, the reference repository for genome-scale metabolic models. A searchable home for curated models, universal metabolite and reaction databases, genomes and compartments, cross-linked to CHEBI, KEGG and MetaCyc. I wrote the pangenome-scale model-generation pipeline behind it.

Open biggr.org Paper in preparation
In-silico twins Live browser

panGEMs Browser

omidard.github.io/panGEMs

A web browser over 2,313 E. coli and 2,346 Lactobacillaceae strain-specific genome-scale metabolic models. Each model is a digital twin of one strain: it predicts what that organism can eat, what it secretes, and which of its genes are essential.

Browse the models Source
Flagship tool Runs in your browser

Flux Studio

omidard.github.io/FluxStudio

A full constraint-based modelling workbench over 4,659 strain models (2,313 E. coli and 2,346 Lactobacillaceae), with no server and no queue: the linear program is solved in the browser tab by a WebAssembly build of GLPK. Thirteen analyses, reaching COBRApy parity for everything a linear solver can do: FBA, pFBA and linear MOMA, loopless CycleFreeFlux, dynamic FBA, flux variability, flux sampling, production envelopes, phenotype phase planes, reaction and gene essentiality (through the GPR rules), synthetic lethality, FSEOF strain design, model QC, multi-model analytics, and cohort comparison. Media come from a curated database of 12,340 growth media, including the chemically defined medium the Lactobacillaceae paper introduced. Every result is validated against COBRApy 0.27, to the fourth decimal.

Launch the studio Source
Models

EcopanGEM

Pangenome-scale reconstruction of E. coli metabolism

Reconstruction and analysis across roughly 12,000 E. coli strains: the largest strain-resolved metabolic model collection assembled for a single bacterial species.

GitHub Browser
ModelsLive browser

LactoPanGEM

Pangenome reconstruction of Lactobacillaceae metabolism

2,346 strain-specific models spanning the whole Lactobacillaceae family, 28 species across 12 genera, predicting species-specific metabolic traits. Published in mSystems.

Browse the models Source
Database

GrowthDB

Measured ground truth for metabolic models

Experimental prokaryote growth rates, uptake and secretion rates, with media and culture conditions, all referenced. This is the measurement a model prediction gets scored against, which is what makes the prediction falsifiable.

GitHub Site
Database

Media

Citation-backed growth media, model-ready

A curated library of growth and simulation media, every component mapped to a standard BiGG exchange reaction and traced to its source. Part of a validation-data suite for genome-scale models.

GitHub Site
Research Science Advances 2026

The genetic basis of metabolic function

First and corresponding author

The same enzyme activity can be encoded by genes that share almost no sequence, and annotation databases quietly paper over this by copying labels between homologs. This work searches millions of genes across thousands of genomes, tests whether each gene is genuinely functional rather than merely annotated, and maps the full space of distinct genetic codes that arrive at one and the same chemistry.

Read the paper Code
Research PLOS Pathogens 2025

Rare genes decide where a strain can live

First author

Reconstructed the panGEM across 12,934 E. coli genomes, validated its gene-knockout predictions to 93% accuracy against experiment, and ran more than 22 million knockout simulations across gut, urine and serum media. 41% of fully sequenced strains carry at least one essential, non-redundant rare gene. The rare tail of the pangenome is not junk. It is a large part of what fits a strain to its niche.

Read the paper Code

02 — Publications

Selected papers and IP

Full list on ORCID.

  1. 2026

    Annotating the pangenome reveals the diversity in the genetic basis for metabolic enzymes

    Science Advances 12(27) · 10.1126/sciadv.aeb3363

    First and corresponding author

  2. 2025

    Rare metabolic gene essentiality is a determinant of microniche adaptation in Escherichia coli

    PLOS Pathogens 21(12) · 10.1371/journal.ppat.1013775 · open code: EcopanGEM

    First author

  3. 2025

    Decomposition of the pangenome matrix reveals a structure in gene distribution in the Escherichia coli species

    mSphere 10(1) · 10.1128/msphere.00532-24

    Second author

  4. 2024

    Pangenome reconstruction of Lactobacillaceae metabolism predicts species-specific metabolic traits

    mSystems 9(7):e00156-24 · 10.1128/msystems.00156-24

    First author

  5. 2021

    Reconstruction and validation of a genome-scale metabolic model of L. lactis NCDO 2118; in-silico succinate and GABA overproduction

    Biochemical Engineering Journal 170:107967 · 10.1016/j.bej.2021.107967

    First author

  6. In
    prep

    BiGG 2026: a pangenome-scale metabolic-model generation pipeline and the next-generation release of the BiGG Models database

    biggr.org

    First author, pipeline architect

  7. Under
    review

    The Escherichia coli pangenome is organized by co-occurring gene sets representing horizontal and vertical inheritance

    Under review at Science Advances

  8. IP

    Lactobacillaceae-related patent

    Main inventor; Notice of Invention filed, DTU.

03 — About

Why I do this

I study bacteria at the scale of whole species rather than single strains, because that is the scale at which the interesting patterns become visible.

One genome tells you what one organism can do. Thousands of them, reconstructed carefully and compared honestly, start to tell you how metabolism is actually built: which genes are load-bearing, how many different ways evolution has found to run the same reaction, and how much of what we call annotation is really a label copied from a neighboring gene. That is the question I have spent most of my career on, and I do not think we are close to done with it.

What has changed lately is not the amount of data but our ability to reason over it. Machine learning, and now language models, let us ask things of these datasets that were previously out of reach: decomposing genomes to find structure nobody thought to look for, or building agents that run a real analysis and then check their own claims against the evidence. Most of what I have found this way was already sitting in public databases. It mainly needed someone with the throughput to go and look.

I did not arrive at computation from computer science. I trained in biology, worked at the bench, and helped take strains from a flask to commercial production before I wrote pipelines for a living. That is the reason I try to keep the modeling honest: a prediction is only worth something if it survives contact with a measurement.

Ph.D. Systems Biology, DTU · trained in the group of Prof. Bernhard O. Palsson

04 — Experience

Where I have worked

  1. 2025 — now

    Postdoctoral Researcher

    DTU Biosustain / DTU BRIGHT, Technical University of Denmark

    Pangenome-scale programs linking sequence and predicted structure to metabolic function; the pipeline behind BiGG 2026; the verification layer the group reuses.

  2. 2025 — now

    Agentic AI systems for biology

    Independent work

    A multi-agent research system on Claude Code that runs real computational biology, built around grounded verifiers: a model does not know when it is wrong, so the system checks the world, not the model.

  3. 2021 — 2024

    Ph.D. Researcher, Systems Biology

    DTU, group of Prof. Bernhard O. Palsson

    Built the E. coli and Lactobacillaceae panGEM collections and the reconstruction pipelines that made that scale possible.

  4. 2019 — 2021

    R&D Specialist, Strain Development

    Behdad Genetics Corp., a biotech startup

    Isolated seven lactic acid bacteria strains and led their scale-up from laboratory to commercial production. Fermentation, QA/QC, GMP and GLP.

Let's talk

Open to work on AI for scientific discovery. If you are building models that need to reason about real biology, I would like to hear from you.

Omidard@biosustain.dtu.dk Download CV