SYBA: Bayesian estimation of synthetic accessibility of organic compounds

Abstract

SYBA (SYnthetic Bayesian Accessibility) is a fragment-based method for the rapid classification of organic compounds as easy-(ES) or hard-to-synthesize (HS). It is based on a Bernoulli naïve Bayes classifier that is used to assign SYBA score contributions to individual fragments based on their frequencies in the database of ES and HS molecules. SYBA was trained on ES molecules available in the ZINC15 database and on HS molecules generated by the Nonpher methodology. SYBA was compared with a random forest, that was utilized as a baseline method, as well as with other two methods for synthetic accessibility assessment: SAScore and SCScore. When used with their suggested thresholds, SYBA improves over random forest classification, albeit marginally, and outperforms SAScore and SCScore. However, upon the optimization of SAScore threshold (that changes from 6.0 to-4.5), SAScore yields similar results as SYBA. Because SYBA is based merely on fragment contributions, it can be used for the analysis of the contribution of individual molecular parts to compound synthetic accessibility. SYBA is publicly available at https ://githu b.com/lich-uct/ syba under the GNU General Public License.

Solvation Free Energy Prediction from Pairwise Atomistic Interactions by Machine Learning

“…Recent progress in machine learning (ML) techniques [55] and their implementation in computational chemistry [75,7] are currently promoting broad applications of SPR in numerous chemical studies [16,27,14,18,56,64,4,1,46,47,52,58,62,24,6,51,57,73,77,2,3,8,9,26,48,54,60,67,72,71,74]. These studies show that ML guarantees faster calculations than computer simulations and more precise estimations than traditional SPR estimations; a considerable number of models showed accuracies comparable to ab initio solvation models in the aqueous system [75].…”

MLSolvA: solvation free energy prediction from pairwise atomistic interactions by machine learning

“…Although we may not expect to obtain detailed chemical or physical insights other than the target property because this is a regression analysis in its nature, SPR has demonstrated significant potential in terms of transferability and outstanding computational efficiency [ 11 , 74 , 79 ]. Recent progress in machine learning (ML) techniques [ 59 ] and their implementation in computational chemistry [ 8 , 79 ] are currently promoting broad applications of SPR in numerous chemical studies [ 1 – 4 , 7 , 9 , 10 , 15 , 18 , 20 , 26 , 28 , 29 , 50 – 52 , 55 , 56 , 58 , 60 – 62 , 64 , 66 , 68 , 71 , 75 – 78 , 82 ]. These studies show that ML guarantees faster calculations than computer simulations and more precise estimations than traditional SPR estimations; a considerable number of models showed accuracies comparable to ab initio solvation models in the aqueous system [ 79 ].…”

LEADD: Lamarckian evolutionary algorithm for de novo drug design

“…Differences between all other pairs of atom typing schemes were of little practical significance (Additional file 1 : Table S4). It’s interesting to note that the mean SAScore values for Morgan atom types fall well below 4.5, which has been suggested as a cut-off for easy to synthesize molecules [ 19 ]. By contrast, the mean SAScore values for dummy and MMFF atom types are approximately 4.6.…”

“…Some solutions revolve around the use of synthetic accessibility (SA) metrics. These metrics may have to be calculated many times throughout the design process, often limiting the user to rather crude rules [ 15 , 16 ] or heuristics [ 17 – 19 ] and precluding the use of more reliable retrosynthetic analyses [ 20 , 21 ]. Post-hoc filtering [ 22 , 23 ], while simple and modular, is computationally inefficient as it might discard solutions in which significant amounts of costs were already sunk.…”