Machine Learning Approach for Prediction of Reaction Yield with Simulated Catalyst Parameters

Abstract

Prediction of reaction yields by machine learning approach is demonstrated in tungsten-catalyzed epoxidation of alkenes. The various electronic and vibrational parameters of the phosphonic acids are collected by DFT simulation, and chosen by LASSO as the essential parameters for prediction of the reaction yields. With the trained model, we can predict yields of the reaction with unverified phosphonic acids with an error of 26%. Keywords: Prediction of reaction yields | Machine learning | Catalyst informaticsRational design and optimization of a catalyst for a transition-metal-catalyzed reaction require the synthesis and examination of various catalyst molecules, but this traditional "trial-and-error" approach generally costs time and labor. Prediction of the reaction yield with machine learning would facilitate the rational design and optimization, and allow discovery of a highly reactive catalyst rapidly and efficiently. Building up this "catalyst informatics" approach, a practical method to develop the prediction model for catalyst reactivity such as the reaction yields is highly desirable.Statistical data analysis has recently been utilized in catalysis for the correlation and prediction of various reaction outcomes such as selectivity and reactivity.1 For example, Sigman et al. have demonstrated multiple linear regression analysis to select the key parameters for the prediction of the enantioselectivities in various transition-metal-catalyzed asymmetric reactions. 2,3They have collected various molecular parameters such as steric, electronic, and vibrational factors by simulation, and have correlated with ¦¦G ‡ values that represent the difference in transition state energies and are proportional to enantiomeric ratio (e.r.). The method has also been applied to the prediction of site-, 4 diastereo-, 5 and chemoselectivities, 6 which are also represented as ¦¦G ‡ values. While various reaction outcomes could be predicted by this methodology, the prediction of the reaction yields in transition-metal-catalyzed reaction is still very rare, 7,8 and a successful example has been recently demonstrated by Doyle et al., who applied the multiple linear regression analysis to correlate the reaction yields of nickel-catalyzed Suzuki coupling reaction with three molecular parameters of phosphine ligands, cone angles (ª), %volume of sphere of radius r occupied by ligand (%V bur ), and electrostatic potential (V min ). 9In this work, we report a new statistical method for the prediction of the reaction yields of the catalytic reaction with logistic regression analysis and LASSO (Least Absolute Shrinkage and Selection Operator).10 Logistic regression analysis is a fitting model that enables development of a suitable model for the percentage values such as the reaction yields compared to the linear regression analysis, and LASSO is a model optimization method that can quantitatively select the key parameters.11 We adapt this approach to predict the reaction yield of the tungsten-catalyzed epoxidation of alkene with hydrogen ...

Computational Ligand Descriptors for Catalyst Design

“…Large database of simple descriptors (topology, size, elements, electronegativity), pruned for high correlation with redox potentials. 15,23,190,191 Bidentate pyrrolide, indolide, aryloxide, and bis(thiolate) ligands (12), applied to analysis of titanium-catalysed hydroamination %Vbur, natural ligand donor parameter (LDP) developed by Odom's group, 192,193 ligand properties derived from simplified, monodentate ligands X on [NCr(NiPr2)2X] 194 Cyclopentadienyl ligands (22) in Rh(III)catalysed C-H activations NMR, CO stretching, redox potential, charges, cone angles, Sterimol parameters 195 P,N-donor and Cp/Cp* ligands (11), coordinated to Ruthenium catalysts for alkene isomerisaton, study of selectivity and activity Initial calculation of 308 descriptors, reduced through further analysis to 6 key descriptors, analysis discussed in section 3.2 below. 196 Asymmetric bidentate ligands (19) with range of donor groups coupled via orthophenylene bridge, coordination to Rh(CO)2 fragments.…”

“…[9][10][11][12][13][14][15] This should not come as a surprise, as it builds on a long tradition of using stereoelectronic parameters, Tolman's perhaps most prominently among them, 16 in this field. 17 Homogeneous catalysis is not (yet) data-rich enough to be considered amenable to "Big Data" approaches, [18][19][20][21] and machine-learning approaches, while used in this area, 9,[22][23][24][25] are still very much in their infancy, 26 but this provides a convenient opportunity to survey and collate available descriptors.…”

Ensemble Learning Approach with LASSO for Predicting Catalytic Reaction Rates

“…For A (i.e., the total yield), a prediction model can be constructed in the same way as in a previous report. 6 In this study, we focused our attention on the construction of a model to predict AB (i.e., the rate constant) using ML from molecular descriptors. Thus, 30 types of descriptor were prepared through DFT calculations to construct the prediction model.…”

“…Of these five, NBO O3 and NBO P are the natural bond orbital values of O 3 and P, I P-O3-H3,b is the infrared vibration intensity of the P-O 3 -H 3 bending mode (Figure 4b), while D O3-P,ss and D H3-O3,s are the fluctuation parameters of the O 3 -P and O 3 -H 3 stretching modes (Figure 4c). 6 The positive coefficient of NBO O3 , which has a negative value, and the negative coefficient of NBO P , which has a positive value, indicate that the reaction rate will increase when their absolute values are relatively close to zero. From a chemical viewpoint, a less negative value of NBO O3 and less positive value of NBO P will reduce the polarization and weaken the strength of the P-O-H bond, leading to enhanced motion in the P-O-H bond.…”

Number Density Descriptor on Extended-Connectivity Fingerprints Combined with Machine Learning Approaches for Predicting Polymer Properties

“…In recent years, researchers have succeeded in applying materials informatics to several functional materials such as thermoelectric materials [3], molecular organic light-emitting diodes [4], and low-thermal-conductivity compounds [5]. In addition, process-structure-property (PSP) linkages [6], Integrated Computational Materials Engineering (ICME) [7], as well as catalyst informatics have been reported recently [8].…”