De la Fuente Lab — Machine Biology Group

Antimicrobial prediction

Our activity predictor reads whole proteomes and finds the encrypted peptides that act as antibiotics.

Anti-inflammatory prediction

Generates peptides scored for both antimicrobial potency and immune modulation.

Secondary structure prediction

Predicts how a peptide folds in the environment it will actually meet, water, membrane or organic solvent, and scores how much its structure shifts between them.

Cell penetration prediction

Predicts cell-penetrating activity across biological membranes to guide intracellular peptide delivery.

Sequence generation from a target profile

A generative language model that writes new peptides from a target profile.

Template-guided optimization

Optimizes a known peptide instead of starting over, searching a generative latent space under a similarity constraint.

Antibiotics for pathogens never seen

Pairs a molecular diffusion model with genome and literature profiles of each strain, predicting activity and generating new molecules for pathogens it has never seen.

Dual-function peptide design

Designs peptides that enter mammalian cells and kill the bacteria hiding inside them.

Reward-guided peptide design

Fine-tunes a 6.4-billion-parameter protein language model with reinforcement learning against predicted MIC.

Sequences cut to fit a fold

Cuts sequences to fit a target backbone, optimizing shape, energy and residue properties with no retraining.