Skip to content
Romero LabDuke University
  • People
  • Publications

Publications

Selected publications.

Complete publication listGoogle Scholar ↗Code, models & resourcesGitHub ↗
01AI that experiments

Learning protein function through autonomous experimental interaction

Demonstrates a new mode of AI learning in which agents discover biological functions and principles through direct experimental interaction.

Brooks C, Notin P, Romero PA

bioRxiv 2026

METL overview combining experimental sequence-function data with simulated biophysical landscapes
02Learning the language of proteins

Biophysics-based protein language models for protein engineering

Shows how AI can learn the physical principles underlying proteins directly from molecular simulation, moving beyond the evolutionary patterns encoded in natural sequences.

Gelman S, Johnson B, Freschlin CR, Sharma A, D’Costa S, Peters J, Gitter A, Romero PA

Nature Methods 2025

Predicted and measured product profiles for engineered olivetolic acid cyclase variants
03Engineering new biological function

Machine learning-guided olivetolic acid cyclase engineering enables tailored cannabinoid biosynthesis in yeast

Shows how engineered enzymes can become programmable control points for directing biological pathways toward desired chemical products.

Blalock N, LaMattina JW, Monge E, Tran R, Louie AE, Urano J, Kambourakis S, Komor RS, Romero PA

bioRxiv 2026

RLXF framework for functionally aligning protein language models with experimental data
04Learning the language of proteins

Functional alignment of protein language models via reinforcement learning

Shows how experimental feedback can align protein language models with engineering goals, enabling them to design functions beyond those explored by natural evolution.

Blalock N, Seshadri S, Nakamura K, Babbar A, Fahlberg SA, Kulkarni A, Romero PA

Nature Communications 2026

SAMPLE self-driving laboratory linking machine learning with automated protein experiments
05AI that experiments

Self-driving laboratories to autonomously navigate the protein fitness landscape

Establishes self-driving laboratories as a new paradigm for autonomous biological engineering.

Rapp JT, Bremer BJ, Romero PA

Nature Chemical Engineering 2024

ACE2 directed-evolution figure showing therapeutic substrates, chromatography, structure, and variant performance
06Engineering new biological function

Directed evolution of angiotensin-converting enzyme 2 peptidase activity profiles for therapeutic applications

Expands the therapeutic potential of ACE2 by engineering its activity and specificity, with potential applications ranging from heart and lung injury to fibrosis, cancer, and neurodegenerative disease.

Heinzelman P, Romero PA

Protein Science 2023

Romero Lab

Duke Biomedical Engineering
Fitzpatrick Center
Duke University
Durham, North Carolina
philip.romero@duke.edu
Google Scholar ↗GitHub ↗Duke BME ↗
© 2026 Romero LabPROTEINS · AI · BIOLOGICAL ENGINEERING