Original title: High-throughput ML-guided design of diverse single-domain antibodies against SARS-CoV-2
Authors: Christof Angermueller,Zelda Marie,Benjamin Jester,Emily Engelhart,Ryan Emerson,Babak Alipanahi,Charles Lin,Colleen Shikany,Daniel Guion,Joel Nelson,Mary Kelley,Margot McMurray,Parker Shaffer,Cameron Cordray,Samer Halabiya,Zachary Ryan McCaw,Sarah Struyvenberg,Kanchan Aggarwal,Stacey Ertel,Anissa Martinez,Snehal Ozarkar,Kevin Hager,Mike Frumkin,Jim Roberts,Randolph M Lopez,David Younger,Lucy Colwell
This article discusses a new approach to treating rapidly evolving pathogenic diseases like COVID-19. The researchers have developed a machine learning-guided sequence design platform that combines experiments and machine learning to create highly diverse antibodies that can bind and neutralize viruses. The key feature of this platform is that it can accurately predict the relationship between antibody sequence and binding activity, even for new viral variants that were not present during training. The researchers discovered that these machine learning-designed antibodies have significant cross-reactivity and can successfully neutralize the Delta and Omicron BA.1 variants of SARS-CoV-2. The antibodies generated by the platform have thousands of variants with improved activity compared to the original sequence, demonstrating the potential for future-proof therapeutics against rapidly evolving pathogens. The authors have stated that they have no competing interests.
Original article: https://www.biorxiv.org/content/10.1101/2023.12.01.569227v2