Background Current AI-based Virtual Screening (VS) methods seek to manage ultra-large molecular libraries. To this end, they develop increasingly efficient heuristics to rank ligands by their predicted activity against a target protein. However, these methods remain computationally demanding due to the billion-scale compound libraries that must be evaluated without prior, informed guidance. Results This article proposes an offline/online method that: (1) Wisely selects (once and forall, offline phase) a small number of easy to compute features <InlineEquation ID="IEq1"><EquationSource Format="TEX">\(F^{}= F^{}_{\text {P}}\cup F^{}_{\text {L}}\)</EquationSource></InlineEquation> of both the amino acid sequence of the proteins (<InlineEquation ID="IEq2"><EquationSource Format="TEX">\(F^{}_{\text {P}}\)</EquationSource></InlineEquation>) and the molecular structure of the ligands (<InlineEquation ID="IEq3"><EquationSource Format="TEX">\(F^{}_{\text {L}}\)</EquationSource></InlineEquation>), and discretises their domains; this induces a low-dimensional finitisation of proteins’ and ligands’ chemical spaces. (2) Given a target protein <InlineEquation ID="IEq4"><EquationSource Format="TEX">\(p\)</EquationSource></InlineEquation>, immediately returns (online phase) a likelihood-based ranking of the classes of the ligands’ chemical space, in descending order of the estimated probability that molecules in each class will achieve a satisfactory activity measurement against <InlineEquation ID="IEq5"><EquationSource Format="TEX">\(p\)</EquationSource></InlineEquation>. This enables any VS method to prioritise the search to the most promising subsets of candidates. To ensure statistically robustness, our offline feature selection: (a) leverages knowledge stemming from a huge dataset of 2 559 403 entries (ligand-protein activity measurements) obtained by unifying the most representative sources regarding biochemical kinetics (Brenda, Sabio-rk, BindingDB) and augmented with 3781 features computed by 7 well-known third-party software tools; (b) explicitly aims at low-dimensional coarse-domain feature spaces; (c) takes proper countermeasures to prevent biases in the source data and overfitting; (d) supports iterative improvement of <InlineEquation ID="IEq6"><EquationSource Format="TEX">\(F^{}\)</EquationSource></InlineEquation> via an anytime offline algorithm and means to interactively exclude features deemed uninformative upon rankings inspection; (e) supports intepretability of the rankings by enabling inspection of the features’ values characterising each ligand class. Conclusions By evaluating our rankings on evaluation data (from PDBbind and additional BindingDB entries unsuitable for accurate statistical analysis), we demonstrate their effectiveness for library prioritisation. Specifically, our findings indicate that approximately 60% of the high-affinity ligands occur in the top 25% ranked ligands’ classes, while 85% fall within the top 50%. Furthermore, we conduct retrospective analysises using AutoDock Vina scores for over 260 000 molecules across 58 medically relevant targets. Results demonstrate that our method cuts the number of dockings needed to retrieve an equivalent set of hits by up to <InlineEquation ID="IEq7"><EquationSource Format="TEX">\(\approx 64\%\)</EquationSource></InlineEquation> on average versus unguided screening.

Prioritising search for virtual screening via preliminary interpretable low-feature likelihood-based rankings of drug-target activity measures

期刊论文
出版模式：

开放获取
发表日期：

2026年7月11日

查看全文

Riccardo Curcio,
Toni Mancini,
Enrico Tronci

摘要

Background

Current AI-based Virtual Screening (VS) methods seek to manage ultra-large molecular libraries. To this end, they develop increasingly efficient heuristics to rank ligands by their predicted activity against a target protein. However, these methods remain computationally demanding due to the billion-scale compound libraries that must be evaluated without prior, informed guidance.

Results

This article proposes an offline/online method that: (1) Wisely selects (once and forall, offline phase) a small number of easy to compute features \(F^{*}= F^{*}_{\text {P}}\cup F^{*}_{\text {L}}\) of both the amino acid sequence of the proteins (\(F^{*}_{\text {P}}\)) and the molecular structure of the ligands (\(F^{*}_{\text {L}}\)), and discretises their domains; this induces a low-dimensional finitisation of proteins’ and ligands’ chemical spaces. (2) Given a target protein \(p\), immediately returns (online phase) a likelihood-based ranking of the classes of the ligands’ chemical space, in descending order of the estimated probability that molecules in each class will achieve a satisfactory activity measurement against \(p\). This enables any VS method to prioritise the search to the most promising subsets of candidates. To ensure statistically robustness, our offline feature selection: (a) leverages knowledge stemming from a huge dataset of 2 559 403 entries (ligand-protein activity measurements) obtained by unifying the most representative sources regarding biochemical kinetics (Brenda, Sabio-rk, BindingDB) and augmented with 3781 features computed by 7 well-known third-party software tools; (b) explicitly aims at low-dimensional coarse-domain feature spaces; (c) takes proper countermeasures to prevent biases in the source data and overfitting; (d) supports iterative improvement of \(F^{*}\) via an anytime offline algorithm and means to interactively exclude features deemed uninformative upon rankings inspection; (e) supports intepretability of the rankings by enabling inspection of the features’ values characterising each ligand class.

Conclusions

By evaluating our rankings on evaluation data (from PDBbind and additional BindingDB entries unsuitable for accurate statistical analysis), we demonstrate their effectiveness for library prioritisation. Specifically, our findings indicate that approximately 60% of the high-affinity ligands occur in the top 25% ranked ligands’ classes, while 85% fall within the top 50%. Furthermore, we conduct retrospective analysises using AutoDock Vina scores for over 260 000 molecules across 58 medically relevant targets. Results demonstrate that our method cuts the number of dockings needed to retrieve an equivalent set of hits by up to \(\approx 64\%\) on average versus unguided screening.