When drug discovery meets web search: Learning to Rank for ligand-based virtual screening

Abstract

BackgroundThe rapid increase in the emergence of novel chemical substances presents a substantial demands for more sophisticated computational methodologies for drug discovery. In this study, the idea of Learning to Rank in web search was presented in drug virtual screening, which has the following unique capabilities of 1). Applicable of identifying compounds on novel targets when there is not enough training data available for these targets, and 2). Integration of heterogeneous data when compound affinities are measured in different platforms.ResultsA standard pipeline was designed to carry out Learning to Rank in virtual screening. Six Learning to Rank algorithms were investigated based on two public datasets collected from Binding Database and the newly-published Community Structure-Activity Resource benchmark dataset. The results have demonstrated that Learning to rank is an efficient computational strategy for drug virtual screening, particularly due to its novel use in cross-target virtual screening and heterogeneous data integration.ConclusionsTo the best of our knowledge, we have introduced here the first application of Learning to Rank in virtual screening. The experiment workflow and algorithm assessment designed in this study will provide a standard protocol for other similar studies. All the datasets as well as the implementations of Learning to Rank algorithms are available at http://www.tongji.edu.cn/~qiliu/lor_vs.html.Graphical AbstractThe analogy between web search and ligand-based drug discovery

PKRank: a novel learning-to-rank method for ligand-based virtual screening using pairwise kernel and RankSVM

“…The ranking prediction model f of learning-to-rank is represented as f ( ) = f ( ( , )) , where ≡ ( , ) is an input feature vector and is a feature map. In this section, we explain the method proposed by Zhang et al that uses the tensor product as [7] as well as the proposed method PKRank, which is a learning-to-rank-based VS using a pairwise kernel. The former method is a special case of the latter, as described presently.…”

“…Rathke et al [6] proposed StructRank, which directly solves the ranking problem and focuses on the Molecular Profiling Research Center for Drug Discovery (molprof), National Institute of Advanced Industrial Science and Technology (AIST), 2-4-7 Aomi, Koto-ku, Tokyo 135-0064, Japan most promising compounds in terms of activity. Zhang et al [7] compared several learning-to-rank prediction models, and concluded that RankSVM is the best. Furthermore, they noted that learning-to-rank can treat multiple heterogeneous experimental data measured for different targets or platforms.…”

“…The purpose of this study is to obtain a more accurate prediction model through PKRank than the method that uses the tensor product [7]. A comparison in terms of prediction accuracy between PKRank and the previous method with compound activity data recorded in BindingDB [10] showed that the former is superior.…”

Acoplamiento molecular basado en ligando por Complejidad LMC

“…La interpretación de los resultados permitió deducir que la mejor aproximación de similaridad entre vectores es con respecto a los vectores activos. los efectos de Maldición de dimensionalidad en el algoritmo propuesto son mínimos debido al empleo de ordenamiento por magnitud de complejidad de los distintos acomplamientos existentes entre Candidatos a medicamento y vectores activos; lo que evita también altos costos computacionales, [23]. Aúnque no es necesario un conjunto de datos entrenamiento como lo exigen los Métodos de aprendizaje supervisado, la precisión del algoritmo propuesto es dificíl de determinar, [9]; pues como Medida de relevancia la Complejidad LMC es un indicador del grado de predicibilidad u orden que sustentan la relación de similaridad entre dos vectores; no se considera un cálculo de error con respecto a un patrón.…”

How Sure Can We Be about ML Methods-Based Evaluation of Compound Activity: Incorporation of Information about Prediction Uncertainty Using Deep Learning Techniques

“…This is the reason why machine learning (ML) methods have recently gained such great popularity in the field of drug design. They are used both to select potential drug candidates from large compounds databases, but also to generate the structures of new chemical compounds de novo-or to optimize their physicochemical and pharmacokinetic properties [14][15][16][17][18][19][20][21][22][23][24][25][26][27].…”