A systematic modeling methodology of deep neural network‐based structure‐property relationship for rapid and reliable prediction on flashpoints

Abstract

Deep neural networks (DNNs) based quantitative structure–property relationship (QSPR) studies are receiving increasing attention due to their excellent performances. A systematic methodology coupling multiple machine learning technologies is proposed to systematically solve vital problems including applicability domain and prediction uncertainty in DNN‐based QSPR modeling. Key features are rapidly extracted from plentiful but chaotic descriptors by principal component analysis (PCA) and kernel PCA. Then, a detailed applicability domain (AD) is defined by K‐means algorithm to avoid unreliable predictions and discover its potential impact on prediction uncertainty. Moreover, prediction uncertainty is analyzed with dropout‐embedded DNN by thousands of independent tests to assess the reliability of predictions. The prediction of flashpoint temperature is employed as a case study, demonstrating that the model accuracy is remarkably improved comparing with the referenced model. Furthermore, the proposed methodology breaks through difficulties in analyzing the uncertainty of DNN‐based QSPRs and presents an AD correlated with the uncertainty.

Treat Molecular Linear Notations as Sentences: Accurate Quantitative Structure–Property Relationship Modeling via a Natural Language Processing Approach

“…In QSPR modeling, the commonly used molecular representations include the qualitative and quantitative molecular descriptors, − molecular graph, − and sequential representation. − Qualitative and quantitative descriptor (QQD)-based QSPR models provide excellent interpretability for two reasons: on one hand, each descriptor has its physical, biological, and chemical meanings; on the other hand, the weight for each descriptor indicates how much it contributes to the target property. However, QQD-based QSPR models have high requirements for feature engineering (i.e., descriptor selection and processing) to reach a decent prediction accuracy.…”

Improved graph‐based multitask learning model with sparse sharing for quantitative structure–property relationship prediction of drug molecules

“…[4][5][6][7][8][9] The quantitative structure-property relationship (QSPR) [10][11][12] refers to a functional relationship that quantitatively predicts the specific properties directly from the molecular structure. [13][14][15][16][17][18] The quality of QSPR models, that is, the evaluation accuracy of the above properties, determines the effectiveness of virtual screening to a significant extent. Therefore, QSPR modeling plays an important role in the pharmaceutical industry.…”

A Transformer-Convolutional Neural Network Based Framework for Predicting Ionic Liquid Properties

“…DNN-based ML systems have aroused great interest by overcoming obstacles of conventional models and obtaining high prediction quality for complex tasks. [35][36][37][38] The growth of deep learning (DL) has provided excellent flexibility and performance to learn molecular fingerprints from data, without explicit guides from experts. [39][40][41] In our previous work, a DNN based recommender system (RS) for predicting the solute-in-IL infinite dilution activity coefficient (γ ∞ ) was developed without any manually designed fingerprint.…”