TY - JOUR
T1 - Computational planning of the synthesis of complex natural products
AU - Mikulak-Klucznik, Barbara
AU - Gołębiowska, Patrycja
AU - Bayly, Alison A.
AU - Popik, Oskar
AU - Klucznik, Tomasz
AU - Szymkuć, Sara
AU - Gajewska, Ewa P.
AU - Dittwald, Piotr
AU - Staszewska-Krajewska, Olga
AU - Beker, Wiktor
AU - Badowski, Tomasz
AU - Scheidt, Karl A.
AU - Molga, Karol
AU - Mlynarski, Jacek
AU - Mrksich, Milan
AU - Grzybowski, Bartosz A.
N1 - Funding Information:
Acknowledgements Development of Chematica was partly supported by US DARPA under the Make-It Award, 69461-CH-DRP #W911NF1610384 (K.M., S.S., E.P.G., P.D., T.B., B.A.G.); the same award also supported the synthesis of dauricine (A.A.B., M.M.). Synthesis of tacamonidine was supported in part (B.M.-K., T.K., B.A.G.) by the National Science Center, NCN, Poland under the Symfonia Award (#2014/12/W/ST5/00592). Synthesis of lamellodysidine A was supported in part (P.G., B.A.G.) by the National Science Center, NCN, Poland under the Maestro Award (#2018/30/A/ST5/00529). J.M. and O.P. thank the Foundation for Polish Science for financial support under award TEAM/2017-4/38. B.A.G. acknowledges support from the Institute for Basic Science Korea, project code IBS-R020-D1. We thank B. Sieredzińska for help in the synthesis of tacamonidine and S. Trice (Merck, KGaA) for help in organizing the Turing test. We thank the following experts for their participation in the Turing test (in alphabetical order): P. Baran (Scripps), J. Bode (ETH Zurich), M. Burke (University of Illinois), M. Christmann (Freie Universität Berlin), H. Davies (Emory University), M. Giedyk (ICHO PAN), D. Huryn (University of Pittsburgh), M. Krische (University of Texas), S. Matsubara (Kyoto University), N. Maulide (Universität Wien), G. Molander (University of Pennsylvania), R. Sarpong (Berkeley), P. Schreiner (Justus Liebig University Giessen) and J. Siitonen (Rice University), as well as four others, who prefer to remain anonymous.
Publisher Copyright:
© 2020, The Author(s), under exclusive licence to Springer Nature Limited.
PY - 2020/12/3
Y1 - 2020/12/3
N2 - Training algorithms to computationally plan multistep organic syntheses has been a challenge for more than 50 years1–7. However, the field has progressed greatly since the development of early programs such as LHASA1,7, for which reaction choices at each step were made by human operators. Multiple software platforms6,8–14 are now capable of completely autonomous planning. But these programs ‘think’ only one step at a time and have so far been limited to relatively simple targets, the syntheses of which could arguably be designed by human chemists within minutes, without the help of a computer. Furthermore, no algorithm has yet been able to design plausible routes to complex natural products, for which much more far-sighted, multistep planning is necessary15,16 and closely related literature precedents cannot be relied on. Here we demonstrate that such computational synthesis planning is possible, provided that the program’s knowledge of organic chemistry and data-based artificial intelligence routines are augmented with causal relationships17,18, allowing it to ‘strategize’ over multiple synthetic steps. Using a Turing-like test administered to synthesis experts, we show that the routes designed by such a program are largely indistinguishable from those designed by humans. We also successfully validated three computer-designed syntheses of natural products in the laboratory. Taken together, these results indicate that expert-level automated synthetic planning is feasible, pending continued improvements to the reaction knowledge base and further code optimization.
AB - Training algorithms to computationally plan multistep organic syntheses has been a challenge for more than 50 years1–7. However, the field has progressed greatly since the development of early programs such as LHASA1,7, for which reaction choices at each step were made by human operators. Multiple software platforms6,8–14 are now capable of completely autonomous planning. But these programs ‘think’ only one step at a time and have so far been limited to relatively simple targets, the syntheses of which could arguably be designed by human chemists within minutes, without the help of a computer. Furthermore, no algorithm has yet been able to design plausible routes to complex natural products, for which much more far-sighted, multistep planning is necessary15,16 and closely related literature precedents cannot be relied on. Here we demonstrate that such computational synthesis planning is possible, provided that the program’s knowledge of organic chemistry and data-based artificial intelligence routines are augmented with causal relationships17,18, allowing it to ‘strategize’ over multiple synthetic steps. Using a Turing-like test administered to synthesis experts, we show that the routes designed by such a program are largely indistinguishable from those designed by humans. We also successfully validated three computer-designed syntheses of natural products in the laboratory. Taken together, these results indicate that expert-level automated synthetic planning is feasible, pending continued improvements to the reaction knowledge base and further code optimization.
UR - http://www.scopus.com/inward/record.url?scp=85092530657&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85092530657&partnerID=8YFLogxK
U2 - 10.1038/s41586-020-2855-y
DO - 10.1038/s41586-020-2855-y
M3 - Article
C2 - 33049755
AN - SCOPUS:85092530657
SN - 0028-0836
VL - 588
SP - 83
EP - 88
JO - Nature
JF - Nature
IS - 7836
ER -