Turning high-throughput structural biology into predictive inhibitor design
Saar KL., McCorkindale W., Fearon D., Boby M., Barr H., Ben-Shmuel A., London N., von Delft F., Chodera JD., Lee AA., Robinson MC., London N., Resnick E., Zaidmann D., Gehrtz P., Reddi RN., Gabizon R., Barr H., Duberstein S., Zidane H., Shurrush K., Cohen G., Solmesky LJ., Lee A., Jajack A., Cvitkovic M., Pan J., Pai R., Ripka EG., Nguyen L., Shafeev M., Matviiuk T., Michurin O., Chernyshenko E., Bilenko VA., Kinakh SO., Logvinenko IG., Melnykov KP., Huliak VD., Tsurupa IS., Gorichko M., Shaikh A., Pinjari J., Swamy V., Pingle M., BVNBS S., Aimon A., Delft FV., Fearon D., Dunnett L., Douangamath A., Dias A., Powell A., Neto JB., Skyner R., Thompson W., Gorrie-Stone T., Walsh M., Owen D., Lukacik P., Strain-Damerell C., Mikolajek H., Horrell S., Koekemoer L., Krojer T., Fairhead M., MacLean EM., Thompson A., Wild CF., Smilova MD., Wright N., Delft AV., Gileadi C., Rangel VL., Schofield C., Salah E., Malla TR., Tumber A., John T., Vakonakis I., Kantsadi AL., Zitzmann N., Brun J., Kiappes JL., Hill M., Witt KD., Alonzi DS., Makower LL., Varghese FS., Overheul GJ., Miesen P., van Rij RP., Jansen J., Smeets B., Tomésio S., Weatherall C., Vaschetto M., Macdonald HB., Chodera JD., Rufa D., Wittmann M., Boby ML., Henry M., Glass WG., Eastman PK., Coffland JE., Dotson DL., Griffen EJ., McCorkindale W., Morris A., Glen R., Cole J., Foster R., Foster H., Calmiano M., Tennant RE., Heer J., Shi J., Jnoff E., Hurley MFD., Lefker BA., Robinson RP., Giroud C., Bennett J., Fedorov O., Reid SP., Morwitzer MJ., Cox L., Morris GM., Ferla M., Moustakas D., Dudgeon T., Pšenák V., Kovar B., Voelz V., Carbery A., Contini A., Clyde A., Ben-Shmuel A., Sittner A., Vitner BPEB., Bar-David E., Tamir H., Achdout H., Levy H., Glinert I., Paran N., Erez N., Puni R., Melamed S., Weiss S., Israely T., Yahalom-Ronen Y., Smalley A., Oleinikovas V., Spencer J., Kenny PW., Ward W., Cattermole E., Ferrins L., Eyermann CJ., Milne BF., Godoy AS., Noske GD., Oliva G., Fernandes RS., Nakamura AM., Gawriljuk VO., White KM., McGovern BL., Rosales R., Garcia-Sastre A., Carney D., Chang E., Saikatendu KS., Neyts LVJ., Donckers K., Jochmans D., Jonghe SD., Bowman GR., Borden B., Singh S., Volkamer A., Rodriguez-Guerra J., Fate G., Hart SH., Bilenko VA., Kinakh SO., Logvinenko IG., Melnykov KP., Huliak VD., Tsurupa IS., Saar KL., Perry B., Fraisse L., Sjö P., Boulet P., Hahn S., Mowbray C., Reid L., Rees P., Huang QYJ., Zvornicanin SN., Shaqra AM., Yilmaz NK., Schiffer CA., Zhang I., Pulido I., Tomlinson C., Taylor JC., Croll TI., Brwewitz L.
A common challenge in drug design pertains to finding chemical modifications to a ligand that increases its affinity to the target protein. An underutilized advance is the increase in structural biology throughput, which has progressed from an artisanal endeavor to a monthly throughput of hundreds of different ligands against a protein in modern synchrotrons. However, the missing piece is a framework that turns high-throughput crystallography data into predictive models for ligand design. Here, we designed a simple machine learning approach that predicts protein–ligand affinity from experimental structures of diverse ligands against a single protein paired with biochemical measurements. Our key insight is using physics-based energy descriptors to represent protein–ligand complexes and a learning-to-rank approach that infers the relevant differences between binding modes. We ran a high-throughput crystallography campaign against the SARS-CoV-2 main protease (MPro), obtaining parallel measurements of over 200 protein–ligand complexes and their binding activities. This allows us to design one-step library syntheses which improved the potency of two distinct micromolar hits by over 10-fold, arriving at a noncovalent and nonpeptidomimetic inhibitor with 120 nM antiviral efficacy. Crucially, our approach successfully extends ligands to unexplored regions of the binding pocket, executing large and fruitful moves in chemical space with simple chemistry.