Molecular Interaction Fingerprints

Rognan, Didier; Desaphy, Jérémy

doi:10.1002/9783527665143.ch14

Cited by 4 publications

(3 citation statements)

References 59 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…12 Post-processing docking poses by alternative scoring schemes has therefore been the subject of intense research during the last decade. 13,14 Notably, machine learning (ML) 15 has gained considerable popularity since the corresponding algorithms (e.g., support vector machines, random forests, decision trees, and deep neural networks) can theoretically delineate subtle nonlinear relationships in a vast hyperparameter space describing experimentally solved protein-ligand complexes. Unfortunately, the true benefit of ML-based SFs remains a matter of debate, 16−18 notably because of a lack of unbiased and shared data/protocols to rigorously compare such methods.…”

Section: ■ Introductionmentioning

confidence: 99%

True Accuracy of Fast Scoring Functions to Predict High-Throughput Screening Data from Docking Poses: The Simpler the Better

Tran-Nguyen

Bret

Rognan

2021

J. Chem. Inf. Model.

Self Cite

View full text Add to dashboard Cite

Hundreds of fast scoring functions have been developed over the last 20 years to predict binding free energies from three-dimensional structures of protein-ligand complexes. Despite numerous statistical promises, we believe that none of them has been properly validated for daily prospective high-throughput virtual screening studies, mostly because in silico screening challenges usually employ artificially built and biased datasets. We here carry out a fully unbiased evaluation of four scoring functions (Pafnucy, ΔvinaRF20, IFP, and GRIM) on an in-house developed data collection of experimental high-confidence screening data (LIT-PCBA) covering about 3 million data points on 15 diverse pharmaceutical targets. All four scoring functions were applied to rescore the docking poses of LIT-PCBA compounds in conditions mimicking exactly standard drug discovery scenarios and were compared in terms of propensity to enrich true binders in the top 1%-ranked hit lists. Interestingly, rescoring based on simple interaction fingerprints or interaction graphs outperforms state-of-the-art machine learning and deep learning scoring functions in most of the cases. The current study notably highlights the strong tendency of deep learning methods to predict affinity values within a very narrow range centered on the mean value of samples used for training. Moreover, it suggests that knowledge of pre-existing binding modes is the key to detecting the most potent binders.

show abstract

Section: ■ Introductionmentioning

confidence: 99%

True Accuracy of Fast Scoring Functions to Predict High-Throughput Screening Data from Docking Poses: The Simpler the Better

Tran-Nguyen

Bret

Rognan

2021

J. Chem. Inf. Model.

Self Cite

View full text Add to dashboard Cite

show abstract

“…In the past, various interaction fingerprints have been developed and successfully employed for post processing of docking poses. For a detailed overview and description of methods, the interested reader may refer to literature such as [ 8 ].…”

Section: Introductionmentioning

confidence: 99%

A novel interaction fingerprint derived from per atom score contributions: exhaustive evaluation of interaction fingerprint performance in docking based virtual screening

et al. 2018

View full text Add to dashboard Cite

Protein ligand interaction fingerprints are a powerful approach for the analysis and assessment of docking poses to improve docking performance in virtual screening. In this study, a novel interaction fingerprint approach (PADIF, protein per atom score contributions derived interaction fingerprint) is presented which was specifically designed for utilising the GOLD scoring functions’ atom contributions together with a specific scoring scheme. This allows the incorporation of known protein–ligand complex structures for a target-specific scoring. Unlike many other methods, this approach uses weighting factors reflecting the relative frequency of a specific interaction in the references and penalizes destabilizing interactions. In addition, and for the first time, an exhaustive validation study was performed that assesses the performance of PADIF and two other interaction fingerprints in virtual screening. Here, PADIF shows superior results, and some rules of thumb for a successful use of interaction fingerprints could be identified.Electronic supplementary materialThe online version of this article (10.1186/s13321-018-0264-0) contains supplementary material, which is available to authorized users.

show abstract

“…GPCR crystal structure-based identification of new ligands with a different functional effect than the co-crystallized ligand has been defined as one of the challenges in the current era of GPCR structural and chemical biology . The growing number of β-adrenoceptor crystal structures meanwhile covers multiple activation states and multiple co-crystallized ligands with different functional efficacies. ,, This unique GPCR structure–function data set allowed us, for the first time, to systematically and separately investigate the effects of both the reference ligands and the receptor conformations in combination with classical (energy-based) , and IFP scoring , approaches on the basis of GPCR crystal structures. In this study, we have investigated whether (and to what extent) (i) the crystal structures retain their preference for ligands with the same functional effect (agonism, antagonism, and inverse agonism) as the co-crystallized ligand, (ii) the different binding site conformations of the crystal structures have an impact on the outcome of VS studies, (iii) specific IFPs can change or amplify the preference of a crystal structure for ligands with a specific functional effect, (iv) the IFP derived from the predicted docking pose of a small agonist (norepinephrine) can be used for the selective retrieval of partial/full agonists (p/fAGO) over antagonist/inverse agonists (ANT/iAGO) and decoys, and (v) sufficient consistent pharmacological data are available to train structure-based VS models to accurately predict biased signaling of GPCR ligands.…”

Section: Introductionmentioning

confidence: 99%

Structure-Based Prediction of G-Protein-Coupled Receptor Ligand Function: A β-Adrenoceptor Case Study

Kooistra

Leurs

Esch

et al. 2015

J. Chem. Inf. Model.

View full text Add to dashboard Cite

The spectacular advances in G-protein-coupled receptor (GPCR) structure determination have opened up new possibilities for structure-based GPCR ligand discovery. The structure-based prediction of whether a ligand stimulates (full/partial agonist), blocks (antagonist), or reduces (inverse agonist) GPCR signaling activity is, however, still challenging. A total of 31 β1 (β1R) and β2 (β2R) adrenoceptor crystal structures, including antagonist, inverse agonist, and partial/full agonist-bound structures, allowed us to explore the possibilities and limitations of structure-based prediction of GPCR ligand function. We used all unique protein-ligand interaction fingerprints (IFPs) derived from all ligand-bound β-adrenergic crystal structure monomers to post-process the docking poses of known β1R/β2R partial/full agonists, antagonists/inverse agonists, and physicochemically similar decoys in each of the β1R/β2R structures. The systematic analysis of these 1920 unique IFP-structure combinations offered new insights into the relative impact of protein conformation and IFP scoring on selective virtual screening (VS) for ligands with a specific functional effect. Our studies show that ligands with the same function can be efficiently classified on the basis of their protein-ligand interaction profile. Small differences between the receptor conformation (used for docking) and reference IFP (used for scoring of the docking poses) determine, however, the enrichment of specific ligand types in VS hit lists. Interestingly, the selective enrichment of partial/full agonists can be achieved by using agonist IFPs to post-process docking poses in agonist-bound as well as antagonist-bound structures. We have identified optimal structure-IFP combinations for the identification and discrimination of antagonists/inverse agonist and partial/full agonists, and defined a predicted IFP for the small full agonist norepinephrine that gave the highest retrieval rate of agonists over antagonists for all structures (with an enrichment factor of 46 for agonists and 8 for antagonists on average at a 1% false-positive rate). This β-adrenoceptor case study provides new insights into the opportunities for selective structure-based discovery of GPCR ligands with a desired function and emphasizes the importance of IFPs in scoring docking poses.

show abstract

Molecular Interaction Fingerprints

Cited by 4 publications

References 59 publications

True Accuracy of Fast Scoring Functions to Predict High-Throughput Screening Data from Docking Poses: The Simpler the Better

True Accuracy of Fast Scoring Functions to Predict High-Throughput Screening Data from Docking Poses: The Simpler the Better

A novel interaction fingerprint derived from per atom score contributions: exhaustive evaluation of interaction fingerprint performance in docking based virtual screening

Structure-Based Prediction of G-Protein-Coupled Receptor Ligand Function: A β-Adrenoceptor Case Study

Contact Info

Product

Resources

About