Francisco Zorrilla

Problem

Reconstructing genome-scale metabolic models from metagenomes was a multi-week, error-prone process that required deep expertise across a dozen separate tools, putting community-level metabolic insight out of reach for most labs.

Approach

Architected and released metaGEM, a Snakemake-orchestrated end-to-end pipeline that automates QC, assembly, binning, MAG consolidation, taxonomy, GEM reconstruction, and community simulation. Designed for HPC clusters and reproducible execution.

Contribution

Lead developer and corresponding author. Designed the architecture, wrote the pipeline, authored the documentation, and have maintained the project through five major releases.

Tools

Attention

Links

Problem

Cheese flavor emerges from cross-feeding between microbial species in the starter culture. Empirical optimization is slow and doesn't generalize across products, so industrial fermenters needed a mechanistic predictor.

Approach

The team (led by first author Chrats Melkonian, Wageningen University) built community-level genome-scale metabolic models for the cheese microbiome, integrated metabolomic and transcriptomic data, and simulated cross-feeding to predict which species pairings drive flavor compounds.

My contribution

Second author. I supported the lead by contributing to the metabolic-modelling component: helping curate models for the relevant species and running community simulations. The wet-lab metabolomics, transcriptomics, and the principal modelling design were driven by Melkonian and the rest of the team.

Tools

Attention

Press coverage

Links

Problem

Plastic pollution outpaces our catalog of enzymes that can degrade it. Wet-lab discovery is slow; we needed a way to mine global metagenomic data for novel plastic-degrading enzyme candidates.

Approach

The team (led by first author Jure Zrimec, Chalmers) built a deep-learning HMM classifier for plastic-degrading enzyme potential, applied it across global ocean and soil metagenomes, and correlated predicted enzyme abundance with regional pollution levels.

My contribution

Co-author. My role was the upstream metagenomic analysis: short-read processing through MAGs, contigs, and abundance estimates that the lead authors then used as input for the HMM-based plastic-degrading enzyme screen.

Tools

Attention

Press coverage

Links

Problem

SynComs for agricultural inoculants and environmental restoration fail unpredictably. A core reason: we don't know which strains are obligately dependent on cross-fed metabolites from their neighbors.

Approach

Built a framework combining genome-scale metabolic modelling, metagenomics, and ecological co-occurrence analysis to flag obligate cross-feeding relationships across soil microbiome datasets.

My contribution

Co-first author. Co-led the metabolic modelling and integration with the metagenomics analysis. The most directly agtech-relevant paper in the portfolio.

Tools

Attention

Links

Problem

OD600 is a universal microbiology readout but conversion to absolute cell count varies wildly between labs and instruments. This is a chronic source of irreproducibility in synthetic biology.

Approach

An iGEM-led consortium coordinated 244 laboratories to test three simple, low-cost OD calibration protocols on eight strains of constitutive GFP-expressing E. coli, establishing a community standard.

My contribution

Co-author. I participated as one of the 244 contributing labs (as a member of the Chalmers iGEM 2018 team), running the wet-lab cultures and OD/flow-cytometry measurements that fed into the consortium dataset.

Tools

Attention

Press coverage

Links

Problem

Antimicrobial drug failure in real microbiomes can't be explained by single-strain resistance alone. Why do communities tolerate drugs better than monocultures?

Approach

The Patil & Ralser teams analysed auxotroph distributions across 12,000+ Earth Microbiome Project communities, then built self-establishing metabolically cooperating yeast communities (SeMeCos) and ran wet-lab drug screens to test the mechanism.

My contribution

Co-author. My specific role was the auxotrophy statistical reanalysis at the start of the paper: re-examining auxotroph distributions across the Earth Microbiome Project, and layering Lisa Maier's drug-on-bug screen data through an auxotroph-vs-prototroph lens to test whether auxotrophic state correlated with drug tolerance.

Tools

Attention

Press coverage

Links

Problem

Enterobacteriaceae bloom in the gut signals dysbiosis and pathogen risk, but the global ecology of which commensal species protect against it was unknown at scale.

Approach

The team re-analysed 12,238 public gut metagenomes spanning 45 countries to identify co-colonisers and co-excluders of Enterobacteriaceae, surfacing a genus-wide colonisation-resistance signal in Faecalibacterium.

My contribution

Co-author. I served as a subject-matter-expert contributor on the metabolic and metagenomic-analysis design, providing guidance during the analysis.

Tools

Attention

Press coverage

Links

Problem

iGEM teams generate mountains of fluorescence data each year. Without consistent calibration, year-over-year results aren't directly comparable.

Approach

The iGEM consortium compared three multi-lab studies measuring fluorescence from identical engineered constructs across years, identifying calibrant preparation as the key driver of inter-year variance.

My contribution

Co-author. As a member of the Chalmers iGEM 2018 team, I contributed wet-lab fluorescence data from our setup; the cross-year analysis and write-up were led by the Beal-coordinated consortium.

Tools

Attention

Press coverage

Links

Problem

Genome-scale metabolic models published in literature often fail to reproduce: only ~40% in BioModels could be re-run without intervention. The community needed a standard.

Approach

The FROG consortium defined and tooled FROG (Flux variability, Reaction deletion, Objective function, Gene deletion) as a reproducibility checkpoint. BioModels integrated it into their submission workflow.

My contribution

Co-author on the consortium paper. My role was contributing model validation results and tool-integration feedback from the metaGEM side; the FROG specification itself was driven by the broader BioModels community.

Tools

Attention

Links

Problem

Hansenula polymorpha (Ogataea polymorpha) is an industrially relevant methylotrophic yeast, but new groups working on it lacked a clear protocol for building a usable GEM.

Approach

Wrote a step-by-step protocol chapter using RAVEN (MATLAB-based reconstruction toolbox) with a homology-based approach. Released the resulting hanpo-GEM publicly.

Contribution

First author of the protocol chapter; built the draft hanpo-GEM as the worked example.

Tools

Attention

Links

Problem

Recurrent C. difficile infection responds to fecal transplant, but the active ingredients of "colonization resistance" are unclear. Which metabolic interactions actually keep the pathogen out?

Approach

The team (Patil Lab, Cambridge) built a synthetic 14-commensal community and probed it against C. difficile invasion to dissect emergent metabolic interactions and their role in colonisation resistance.

My contribution

Co-author. I reconstructed and simulated GEMs from context-specific genomes, ran the metabolomics analysis between suppressive and non-suppressive samples, and identified genomic signatures consistent with the observed metabolomic differences.

Tools

Attention

Links

Scope

This thesis sits at the intersection of microbial ecology, metagenomics, and metabolic modelling, outlining the development and applications of omics-driven metabolic-modelling approaches to understand microbial community metabolism in diverse ecological contexts.

Methodological backbone

I developed metaGEM, a workflow for reconstructing context-specific genome-scale metabolic models (GEMs) from metagenome-assembled genomes (MAGs) and predicting nutritional dependencies directly from shotgun metagenomes. I applied it across five biomes (synthetic lab cultures, human gut, plant-associated, bulk soil, and ocean metagenomes), reconstructing over 14,000 MAGs and corresponding GEMs.

Three applied case studies

• Auxotrophy distributions across microbiomes: high amino-acid auxotroph frequency, higher in host-associated than free-living samples, with evidence that auxotrophy correlates with drug tolerance. A second collaboration showed that purely genome-based annotations over-estimate auxotrophies while metagenomic models under-estimate them, with insertion-sequence enrichment in auxotrophic genomes hinting at a gene-loss mechanism.
• Cheese-flavour formation (industrial collaboration): metabolic models from fermentation-culture genomes, simulated under varying media, revealed amino-acid auxotrophies in three community members rescued by a fourth, and strain-specific contributions to flavour.
• C. difficile invasion resistance: GEMs from a defined 14-member community, plus metabolomics and genomic-signature analysis, to dissect why some sub-communities suppress the pathogen.

gutDBX

A separate methodological chapter introduces gutDBX: a metagenomics-driven database of 141,556 non-dereplicated microbial sequences associated with xenobiotic and pharmacologically-relevant metabolism in human and mouse gut, validated against the MetaCardis cohort. Hits correlated with serum cholesterol, Shannon diversity, and total prescribed-drug count.

Outlook

The closing chapter discusses future directions for metaGEM, including global plastic-degrading-potential surveys, long-read sequencing of understudied environments, and paleometagenomics.

Tools

Links