Horizontal Gene Transfer (HGT) is a process where genes are transferred across species, rather than inherited vertically. There are two main ways to detect it: the first focuses on single genomes, looking for unusual patterns in their DNA, such as GC content outliers—genomic “footprints” that suggest foreign genetic material. The second compares multiple genomes by reconstructing gene and species trees. Discrepancies between these trees often reveal candidate transfer events. The research projects below explore this second approach, using theoretical models and algorithms to detect and explain horizontal transfers from evolutionary data. 

Perfect Transfer Networks

Project type: Research · Theoretical Bioinformatics · Phylogenetic Networks · Algorithm Design 

Role: Lead author · Model definition · Algorithm design · Theoretical Analysis· Communication

Associated paper:

López Sánchez, A. and Lafond, M.

Predicting horizontal gene transfers with perfect transfer networks, Algorithms for Molecular Biology, 2024.

DOI: 10.1186/s13015-023-00242-2 

Published: Peer-reviewed journal · Algorithms for Molecular Biology (Springer) 

The Challenge

This was my first PhD project, and the seed of a new research direction. We asked whether it's possible to predict horizontal gene transfers using traits rather than sequences—focusing on persistent characters like biochemical markers or transposable elements that are rarely lost once gained. Inspired by classical models like perfect phylogenies and recombination networks, I introduced the concept of Perfect Transfer Networks (PTNs): phylogenetic networks that explain character evolution under minimal loss and unique origin assumptions. 

Contributions

This project was not only foundational for my PhD, but also widely presented—from academic conferences to general audiences via the 3MT contest and Google Developer Group events. 

A Sankoff-Rousseau-like Algorithm for Minimizing Lateral Gene Transfers and Losses 

Project type: Research · Theoretical Bioinformatics · Phylogenetic Networks · Algorithm Design 

Role: Lead author · Model definition · Algorithm design · International Collaboration

Presented at:
RECOMB-CG 2025 · Seoul, South Korea
22nd International Conference on Comparative Genomics, co-located with RECOMB
Proceedings will be published by Springer Nature
Proceedings Link

The Challenge

This project extended the Perfect Transfer Network framework to allow for gene loss events, addressing a key limitation of earlier models. It began after I presented my PTN work and was contacted about a related open question—this sparked an international collaboration and led to two research visits to the University of Leipzig’s Bioinformatics group.

At the heart of the project is a dynamic programming algorithm—akin to the classical Sankoff-Rousseau method—designed to infer the simplest evolutionary scenarios involving both horizontal gene transfers and losses, under a single-origin assumption. The algorithm balances parsimony and biological realism by letting users assign different penalties to transfers and losses.

This project also marked a personal milestone—it helped me break through the conceptual barrier of dynamic programming and apply it meaningfully to biological data.