Fingerprint‐based computational models of 5‐lipo‐oxygenase activating protein inhibitors: Activity prediction and structure clustering

Abstract

Inflammatory diseases can be treated by inhibiting 5-lipo-oxygenase activating protein (FLAP). In this study, a data set containing 2,112 FLAP inhibitors was collected. A total of 25 classification models were built by five machine learning algorithms with five different types of fingerprints. The best model, which was built by support vector machine algorithm with ECFP_4 fingerprint had an accuracy and a Matthews correlation coefficient of 0.862 and 0.722 on the test set, respectively. The predicted results were further evaluated by the application domain d STD-PRO (a distance between one compound to models). Each compound had a d STD-PRO value, which was calculated by the predicted probabilities obtained from all 25 models. The application domain results suggested that the reliability of predicted results depended mainly on the compounds themselves rather than algorithms or fingerprints. A group of customized 10-bit fingerprint was manually defined for clustering the molecular structures of 2,112 FLAP inhibitors into eight subsets by K-Means. According to the clustering results, most of inhibitors in two subsets (subsets 2 and 4) were highly active inhibitors. We found that aryl oxadiazole/oxazole alkanes, biaryl amino-heteroarenes, two aromatic rings (often N-containing) linked by a cyclobutene group, and 1,2,4-triazole group were typical fragments in highly active inhibitors. K E Y W O R D S 5-lipo-oxygenase activating protein inhibitor, classification model, machine learning, structure clustering, support vector machine 932 | TU eT al.

Attention Mechanism-Based Graph Neural Network Model for Effective Activity Prediction of SARS-CoV-2 Main Protease Inhibitors: Application to Drug Repurposing as Potential COVID-19 Therapy

“…Currently, research has been proposed to combine GNN features with different ML algorithms for the classification or regression tasks. Besides, ECFP fingerprints were also extensively used for the representation of molecules because of their easier interpretation and faster computation compared to other descriptors, and moreover, they have been successfully applied in various prediction tasks. − Here, four classic supervised learners were also considered to be coupled with GNN neural features and ECFP fingerprints, including SVR, RF, XGBoost, and LightGBM. Since GAT-10 yields the optimal performance among 36 GNN models, the neural features extracted by GAT-10 were used in the subsequent comparative study.…”

Combined Machine Learning and GRID-Independent Molecular Descriptor (GRIND) Models to Probe the Activity Profiles of 5-Lipoxygenase Activating Protein Inhibitors

“…In terms of the best model, XGBoost outperformed all and the accuracy and MCC values were observed as 0.90 and 0.81, respectively. Additionally, a pervious fingerprint-based ML study on FLAP modulators stated that that the reliability of predicted results depends mainly on the compounds themselves rather than algorithms or fingerprints ( Tu et al, 2020 ).…”

Prediction and Structure–Activity Relationship Analysis on Ready Biodegradability of Chemical Using Machine Learning Method

“…The formula is where y ( J ) and σ­( J ) are the average and standard deviation of the probability of n models predicted for compound J , respectively. N ( x , y ( J ), σ­( J )) is the normal distribution density function of probability …”

“…N(x, y(J), σ(J)) is the normal distribution density function of probability. 51 On the basis of the prediction accuracy and standard deviation of multiple models for the same compound, we can obtain a normal distribution curve. The d STD-PRO value represents the area under the curve and also the probability when the compound is predicted to be of the opposite type.…”