<scp>ClassicalGSG</scp>: Prediction of <scp>log</scp><i>P</i> using classical molecular force fields and geometric scattering for graphs

Donyapour, Nazanin; Hirn, Matthew; Dickson, Alex

doi:10.1002/jcc.26519

Cited by 8 publications

(18 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Submission ClassicalGSG DB3 is an empirical method that employed neural networks (NNs) where the inputs are molecular features generated using a method called Geometric Scattering for Graphs (GSG) [84][85][86]. In GSG, atomic features are transformed into molecular features using the graph molecular structure.…”

Section: A Shortlist Of Consistently Well-performing Methods In the Log P Challengementioning

confidence: 99%

Evaluation of log P, pKa, and log D predictions from the SAMPL7 blind challenge

Bergazin

Tielker

Zhang

et al. 2021

J Comput Aided Mol Des

View full text Add to dashboard Cite

The Statistical Assessment of Modeling of Proteins and Ligands (SAMPL) challenges focuses the computational modeling community on areas in need of improvement for rational drug design. The SAMPL7 physical property challenge dealt with prediction of octanol-water partition coefficients and pKa for 22 compounds. The dataset was composed of a series of N-acylsulfonamides and related bioisosteres. 17 research groups participated in the log P challenge, submitting 33 blind submissions total. For the pKa challenge, 7 different groups participated, submitting 9 blind submissions in total. Overall, the accuracy of octanol-water log P predictions in the SAMPL7 challenge was lower than octanol-water log P predictions in SAMPL6, likely due to a more diverse dataset. Compared to the SAMPL6 pKa challenge, accuracy remains unchanged in SAMPL7. Interestingly, here, though macroscopic pKa values were often predicted with reasonable accuracy, there was dramatically more disagreement among participants as to which microscopic transitions produced these values (with methods often disagreeing even as to the sign of the free energy change associated with certain transitions), indicating far more work needs to be done on pKa prediction methods.

show abstract

Section: A Shortlist Of Consistently Well-performing Methods In the Log P Challengementioning

confidence: 99%

Evaluation of log P, pKa, and log D predictions from the SAMPL7 blind challenge

Bergazin

Tielker

Zhang

et al. 2021

J Comput Aided Mol Des

View full text Add to dashboard Cite

show abstract

“…One of the main challenges of this project was finding a representation that could describe molecular topologies uniquely. For this purpose, we originally tried features from a method called "Geometric Scattering for Graphs" [83], which we had used previously to predict partition coefficients for small molecules [84,85]. We attempted to use these features by themselves, as well as a combination of the GSG features with the Behler-Parrinello symmetry functions used above, and none of these feature sets were able to generate a proper models for FT simulations.…”

Section: Discussionmentioning

confidence: 99%

Flexible Topology: A New Method for Dynamic Drug Design

Donyapour

Roussey

Bose

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Ligand-induced conformational changes form the underpinnings of most essential biomolecular processes, however they are often neglected in screening and design applications, due to the high computational cost. We propose a method called “Flexible Topology”, where a ligand is comprised of a set of shapeshifting “ghost” atoms, whose atomic identities and connectivity can dynamically change over the course of a simulation. Ghost atoms are guided toward their target positions using a translation-, rotation-, and index-invariant restraint potential. When implemented with a large set possible targets, this can simulate trajectories in a coupled chemical-conformational space that allows solutions to molecular design problems to simply emerge over the course of a trajectory. It also provides a mechanism for observing ligand-induced conformational change, by allowing for the surrounding environment and the ghost atoms to respond to each other during the design process. This builds on a substantial history of alchemy in the field of molecular dynamics simulation, including the Lambda dynamics method developed by Brooks and coworkers [X. Kong and C.L. Brooks III, J. Chem. Phys. 105, 2414 (1996)], but takes it to an extreme by associating a set of four dynamical variables with each shapeshifting atom that control not only its presence but its atomic identity. Here we outline the theoretical details of this method, its implementation using the OpenMM simulation package, and some preliminary studies of ghost particle assembly simulations. We show that while Flexible Topology is able to consistently assemble molecules up to 50 atoms in size, modifications are needed before it can reliably predict bound poses in heterogeneous environments.

show abstract

“…Considering the widespread usage of log P and the cost associated with experimental measurements, a large variety of computational methods such as XlogP3 [ 8 ], AlogP [ 9 ], ClogP [ 10 ], KowWIN [ 11 ], JPlogP [ 12 ] László et al [ 13 ], Huuskonen et al [ 14 ], MlogP [ 15 ], iLogP [ 16 ], Manhold [ 17 ], AlogPS [ 18 ], S+logP [ 19 ], CSLogP [ 20 ], Silicos-IT LogP [ 21 ], TopP-S [ 22 ], OpenChem [ 23 ] have been developed over the years. These methods employ various techniques and algorithms for predicting log P and have their pros and cons as explained in our previous work [ 24 ]. In publications, these methods often use their own specific test sets, making the comparison between different algorithms challenging.…”

Section: Introductionmentioning

confidence: 99%

“…The submitted prediction methods are classified into one of the following categories: Empirical methods, Physical molecular mechanics (MM)-based, Physical quantum mechanics (QM)-based, or Mixed methods. Empirical methods [ 8 – 22 , 24 , 28 , 29 ] are data-driven methods in which predictor models are trained directly on a dataset of molecules. The empirical category includes methods that employ atomic/fragment-based additive methods, machine learning, and quantitative structure-property relationship (QSPR) approaches.…”

Section: Introductionmentioning

confidence: 99%

“…In our previous work [ 24 ], we employed ClassicalGSG to examine the performance of two force field parameter matching programs: CHARMM General Force Field (CGenFF) [ 54 , 55 ] and General AMBER Force Field 2 (GAFF2) [ 56 , 57 ]. The NN models were trained using a dataset of molecules made available by OpenChem [ 23 ] and we showed that CGenFF-generated parameters with a specific ad hoc scheme of classifying CGenFF atomic types achieved the highest accuracy in predicting log P values.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Predicting partition coefficients for the SAMPL7 physical property challenge using the ClassicalGSG method

Donyapour

Dickson

2021

J Comput Aided Mol Des

Self Cite

View full text Add to dashboard Cite

The prediction of log P values is one part of the statistical assessment of the modeling of proteins and ligands (SAMPL) blind challenges. Here, we use a molecular graph representation method called Geometric Scattering for Graphs (GSG) to transform atomic attributes to molecular features. The atomic attributes used here are parameters from classical molecular force fields including partial charges and Lennard-Jones interaction parameters. The molecular features from GSG are used as inputs to neural networks that are trained using a “master” dataset comprised of over 41, 000 unique log P values. The specific molecular targets in the SAMPL7 log P prediction challenge were unique in that they all contained a sulfonyl moeity. This motivated a set of ClassicalGSG submissions where predictors were trained on different subsets of the master dataset that are filtered according to chemical types and/or the presence of the sulfonyl moeity. We find that our ranked prediction obtained 5th place with an RMSE of 0.77 log P units and an MAE of 0.62, while one of our non-ranked predictions achieved first place among all submissions with an RMSE of 0.55 and an MAE of 0.44. After the conclusion of the challenge we also examined the performance of open-source force field parameters that allow for an end-to-end log P predictor model: General AMBER Force Field (GAFF), Universal Force Field (UFF), Merck Molecular Force Field 94 (MMFF94) and Ghemical. We find that ClassicalGSG models trained with atomic attributes from MMFF94 can yield more accurate predictions compared to those trained with CGenFF atomic attributes.

show abstract

ClassicalGSG: Prediction of logP using classical molecular force fields and geometric scattering for graphs

Cited by 8 publications

References 50 publications

Evaluation of log P, pKa, and log D predictions from the SAMPL7 blind challenge

Evaluation of log P, pKa, and log D predictions from the SAMPL7 blind challenge

Flexible Topology: A New Method for Dynamic Drug Design

Predicting partition coefficients for the SAMPL7 physical property challenge using the ClassicalGSG method

Contact Info

Product

Resources

About