I’m Fei, a Principal Scientist at Lila Sciences, where I develop machine learning methods to design proteins, mRNA, and DNA regulatory elements for therapeutic applications.
I got my PhD in Physics from the Niels Bohr Institute at the University of Copenhagen, advised by Kresten Lindorff-Larsen, Wouter Boomsma, Mogens Høgh Jensen and Daniel Otzen, where I built computational algorithms to predict protein structure and dynamics. My work combined dynamic Bayesian networks and Markov chain Monte Carlo simulations to capture protein conformational fluctuations [Ref 1]. I also accurately predicted the structure of an amyloid protein by decoding evolutionary patterns in protein sequences [Ref 2], 5 years before AlphaFold’s release.
During my postdoc with Robert Best at the NIH, we discovered that generative model capturing covariant patterns within a protein family can be used effectively to design novel protein sequences. In 2018, we published (arguably) the first study applying a generative model to create high-diversity protein sequences, validated through wet-lab experiments [Ref 3]. This was followed by a subsequent study on a larger protein [Ref 4]. In addition to protein design, I also developed molecular dynamics simulation methods that explain how proteins fold on the ribosome [Ref 5–8].
Afterwards, at Novoyzmes R&D, I applied my research to design several enzyme products for the global market, including Pristine®, Carezyme®, Mannaway®, and others.
References
[1] Equilibrium simulations of proteins using molecular fragment replacement and NMR chemical shifts. PNAS, 2014.
[2] Structure of a functional amyloid protein subunit computed using sequence variation. JACS, 2015.
[3] Co‐evolutionary fitness landscapes for sequence design. Angewandte Chemie, 2018.
[4] Design of a protein with improved thermal stability by an evolution‐based generative model. Angewandte Chemie, 2022.
[5] The folding pathway of an Ig domain is conserved on and off the ribosome. PNAS, 2018.
[6] A small single-domain protein folds through the same pathway on- and off-the ribosome. PNAS, 2018.
[7] The Shape of the ribosome exit tunnel affects cotranslational protein folding. eLife, 2018.
[8] Cotranslational folding cooperativity of contiguous domains of α-spectrin. PNAS, 2020.