George M. Church, PhD, Robert Winthrop Professor, Genetics, Harvard Medical School
In 1974, I adapted energy optimization methods for use in models of nucleic acids, protein and their interactions, and then for use in crystallographic refinement. In the last days of the second millennium,
David Baker's team won the Critical Assessment of Structure Prediction (CASP) by an unbelievable margin. Since then, our labs exchanged 3 Ph.D. students (Dantas, Raman, Lajoie), Wannier from Mayo's group, Stranges from Kuhlman, and Mandell from Kortemme. We engineered new sensor proteins for metabolic
engineering, essential proteins with non-standard amino acids for biocontainment, and polymerase-pore fusions
for nanopore sequencing. None of this prepared us for the revolution following Gleb Kuznetsov joining our lab in 2012, joined soon by Surge Biswas, Pierce Ogden, Ethan Alley, and Sam Sinai. Together we abruptly moved to "sequence-only", deep machine learning for protein design -- ranging from fluorescent proteins to AAV capsids to antibodies. When combined
with libraries of millions of designed gene segments from chip-synthesis and rapid testing, each design cycle can take large leaps in sequence space and function space.