Self-Organizing Maps in Drug Discovery: Compound Library Design, Scaffold-Hopping, Repurposing

Schneider, Petra; Tanrıkulu, Yusuf; Schneider, Gisbert

doi:10.2174/092986709787002655

Cited by 114 publications

(63 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…2A). SOMs were originally developed as a neural network-inspired heuristic to reduce dimensionality (9), and they have become a workhorse of molecular informatics (31,32). Using the unsupervised SOM algorithm, we clustered 12,661 manually annotated, pharmaceutically relevant drugs and lead compounds [collection of bioactive reference analogs (COBRA); inSili.com LLC] (33).…”

Section: Resultsmentioning

confidence: 99%

Identifying the macromolecular targets of de novo-designed chemical entities through self-organizing map consensus

Reker

Rodrigues

Schneider

2014

Proc. Natl. Acad. Sci. U.S.A.

210

165

View full text Add to dashboard Cite

De novo molecular design and in silico prediction of polypharmacological profiles are emerging research topics that will profoundly affect the future of drug discovery and chemical biology. The goal is to identify the macromolecular targets of new chemical agents. Although several computational tools for predicting such targets are publicly available, none of these methods was explicitly designed to predict target engagement by de novo-designed molecules. Here we present the development and practical application of a unique technique, self-organizing map-based prediction of drug equivalence relationships (SPiDER), that merges the concepts of self-organizing maps, consensus scoring, and statistical analysis to successfully identify targets for both known drugs and computer-generated molecular scaffolds. We discovered a potential off-target liability of fenofibraterelated compounds, and in a comprehensive prospective application, we identified a multitarget-modulating profile of de novo designed molecules. These results demonstrate that SPiDER may be used to identify innovative compounds in chemical biology and in the early stages of drug discovery, and help investigate the potential side effects of drugs and their repurposing options.drug design | target prediction | polypharmacology | machine learning | chemical similarity C omputer-assisted de novo molecular design has evolved as a popular source of ideas to combat the perceived lack of new chemical entities (NCEs) in chemical biology and drug discovery (1). We demonstrate that automated de novo design delivers readily synthesizable NCEs with desirable activity profiles. Although receptor-based design operates on a model of a macromolecular binding site, ligand-based methods are either explicitly or implicitly based on the chemical similarity principle (2) without requiring a receptor model (3). Instead, the latter typically uses some measure of chemical or pharmacophore feature similarity to a reference ligand as a fitness function, which aims to generate NCEs as template mimetics via scaffold hopping (4-8). We report the development, implementation, and successful prospective application of an innovative computational technique for the target profiling of de novo-designed molecules. The approach combines the concepts of self-organizing maps (SOMs) (9) for macromolecular target prediction (10), consensus scoring (11), and a statistical evaluation that provides confidence estimates for the predictions.Predicting polypharmacological activities is a topic relevant to chemical biology and drug discovery, not only to take advantage of inherent drug promiscuity but also to decrease lead compound attrition caused by unfavorable off-target modulation

show abstract

Section: Resultsmentioning

confidence: 99%

Identifying the macromolecular targets of de novo-designed chemical entities through self-organizing map consensus

Reker

Rodrigues

Schneider

2014

Proc. Natl. Acad. Sci. U.S.A.

210

165

View full text Add to dashboard Cite

show abstract

“…The purpose of chemogenomics is to find any interesting inhibitors against an orphan target receptor whose function has not been understood and its inhibitor has not been known [22]. Schneider et al have applied SOM for mapping different ligand classes and succesfully selected a promising library from many vendor catalogues [23]. SOMPLS is considered to be the advanced method from their approach.…”

Section: Resultsmentioning

confidence: 99%

Visualization and Chemical Interpretation of Multi-Target Structure-Activity Relationships Using SOMPLS

長谷川¹,

船津²

2011

J. Comput. Aided Chem.

View full text Add to dashboard Cite

In quantitative structure-activity relationships (QSAR), partial least squares (PLS) are of particular interest as a statistical method. Since successful applications of PLS to QSAR data set, PLS has evolved for coping with more demands associated with complex data structures. Especially, PLS variants focusing on visualization and chemical interpretation are highly desirable in modeling multi-target structure-activity relationships. In this paper, we employed the self-organized PLS (SOMPLS) approach to predict multiple inhibitory activities against three serine protease receptors (Thrombin, Trypsin and Factor Xa). Volsurf descriptors were used as chemical descriptors.

show abstract

“…On the one hand, unsupervised mapping methods, such as Principal Coordinate Analysis (PCA) or Self-Organizing Maps (SOM), have been broadly used in chemoinformatics. [5,[19][20][21] These methods do not take into account target values during model generation. On the other hand, supervised or targetdriven subspace mapping methods have not been much exploited in the literature so far, with the exception of Partial Least Squares (PLS) and Linear Discriminant Analysis (LDA) that fall into this category.…”

Section: Related Workmentioning

confidence: 99%

Target‐Driven Subspace Mapping Methods and Their Applicability Domain Estimation

Soto

Vazquez

Strickert

et al. 2011

Molecular Informatics

View full text Add to dashboard Cite

This work describes a methodology for assisting virtual screening of drugs during the early stages of the drug development process. This methodology is proposed to improve the reliability of in silico property prediction and it is structured in two steps. Firstly, a transformation is sought for mapping a high-dimensional space defined by potentially redundant or irrelevant molecular descriptors into a low-dimensional application-related space. For this task we evaluate three different target-driven subspace mapping methods, out of which we highlight the recent Correlative Matrix Mapping (CMM) as the most stable. Secondly, we apply an applicability domain model on the low-dimensional space for assessing confidentiality of compound classification. By a probabilistic framework the applicability domain approach identifies poorly represented compounds in the training set (extrapolation problems) and regions in the space where the uncertainty about the correct class is higher than normal (interpolation problems). This two-step approach represents an important contribution to the development of confident prediction tools in the chemoinformatics area, where the field is in need of both interpretable models and methods that estimate the confidence of predictions.

show abstract

Self-Organizing Maps in Drug Discovery: Compound Library Design, Scaffold-Hopping, Repurposing

Cited by 114 publications

References 47 publications

Identifying the macromolecular targets of de novo-designed chemical entities through self-organizing map consensus

Identifying the macromolecular targets of de novo-designed chemical entities through self-organizing map consensus

Visualization and Chemical Interpretation of Multi-Target Structure-Activity Relationships Using SOMPLS

Target‐Driven Subspace Mapping Methods and Their Applicability Domain Estimation

Contact Info

Product

Resources

About