Predicting reaction performance in C–N cross-coupling using machine learning

Abstract

Machine learning methods are becoming integral to scientific inquiry in numerous disciplines. We demonstrated that machine learning can be used to predict the performance of a synthetic reaction in multidimensional chemical space using data obtained via high-throughput experimentation. We created scripts to compute and extract atomic, molecular, and vibrational descriptors for the components of a palladium-catalyzed Buchwald-Hartwig cross-coupling of aryl halides with 4-methylaniline in the presence of various potentially inhibitory additives. Using these descriptors as inputs and reaction yield as output, we showed that a random forest algorithm provides significantly improved predictive performance over linear regression analysis. The random forest model was also successfully applied to sparse training sets and out-of-sample prediction, suggesting its value in facilitating adoption of synthetic methodology.

Gold‐Catalyzed Cross‐Coupling Reactions: An Overview of Design Strategies, Mechanistic Studies, and Applications

“…However,f uture research should focus on developing selectivea nd scalable reactions that can proceed catalytically under ambient conditions and without the requirements of sacrificialo xidants and stoichiometric amounts of expensive reagents. [422,423] Traditionalt ransition-metal catalysis including crosscoupling reactions [424,425] are already being tested and advancedb ya pplying these modernc omputational tools. The substrate scope for dual Au and visible-light-mediated reactions is ratherl imited because most methodologies require the use of ad iazonium salt as ac oupling partner.S imilarly,o nly af ew examples are availablef or the use of Pd-free plasmonic catalysis by Au for CÀCc oupling reactions.…”

Rapid Assessment of the Reaction‐Condition‐Based Sensitivity of Chemical Transformations

“…[9] As photochemistry is not yet fully established within the mainstream synthetic community and fundamentally very different setups are available,there is an eed to ensure confidence in photochemistry.Asimpler approach could allow the parameters of existing setups to be more clearly defined and controlled, two issues known to hamper reproducibility. [12] Finally,s uch data presentation can be useful to industrial applications as ab asis for further "design of experiment" (DOE) optimization, as knowledge about the most influential parameters accelerates the selection of variables screened under DOE. Taking the underrepresentation of photochemistry caused by reproducibility issues as an example,w es ought to develop ag eneral tool to support the reproducibility of reactions.W eenvisaged that this could be achieved by evaluating the influence of key reaction parameters,a nd that this tool would complement standard deviation and optimization tables,a nd additionally increase attention by presenting the data in ar eadily comprehensible manner.I np articular, the intuitive presentation of am ethodsa ir and water sensitivity would be key to promoting its use.Inaddition to building confidence in anew method, this tool could also provide help for reaction troubleshooting.Furthermore,the collection and publication of complete and standardized data is of paramount importance to data scientists in order to identify correlations,create predictive algorithms,a nd support machine learning projects.…”

Hochdurchsatzstrategien zur Entdeckung und Optimierung katalytischer Reaktionen

“…[12,47] Beispielsweise schrieben Doyle,D reher und Mitarbeiter ein frei zugängliches Computerskript zum Erhalt zahlreicher potenziell wichtiger chemischer Deskriptoren (z.B.M olekülvolumen, E HOMO ,E lektronegativität) aus manuell in Spartan gezeichneten Strukturen. [48] Mit diesen nummerischen Deskriptoren als Eingabe und den in der HTE als Ausgabe erstellten prozentualen Ausbeuten konnten Algorithmen des maschinellen Lernens bei der Suche nach Vorhersagemodellen eingesetzt werden. Zwar müssen solche Strategien mit Vorsicht umgesetzt werden, [49] das Potenzial von statistischen Analysemethoden zur Aufdeckung von Logik und zur Vorhersage von Reaktionen mit nicht-intuitivem Verhalten ist indes enorm.…”