Browse
Submit Data
Databases
API
Help

Dataset Information

0 Views

0 Connections

0 Citations

0 Reanalyses

0 Downloads

Omics score: 0

Data Set Augmentation Allows Deep Learning-Based Virtual Screening to Better Generalize to Unseen Target Classes and Highlight Important Binding Interactions.

ABSTRACT: Current deep learning methods for structure-based virtual screening take the structures of both the protein and the ligand as input but make little or no use of the protein structure when predicting ligand binding. Here, we show how a relatively simple method of data set augmentation forces such deep learning methods to take into account information from the protein. Models trained in this way are more generalizable (make better predictions on protein/ligand complexes from a different distribution to the training data). They also assign more meaningful importance to the protein and ligand atoms involved in binding. Overall, our results show that data set augmentation can help deep learning-based virtual screening to learn physical interactions rather than data set biases.

SUBMITTER: Scantlebury J

PROVIDER: S-EPMC7611237 | biostudies-literature | 2020 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Json Xml

Publications

Data Set Augmentation Allows Deep Learning-Based Virtual Screening to Better Generalize to Unseen Target Classes and Highlight Important Binding Interactions.

Scantlebury Jack J Brown Nathan N Von Delft Frank F Deane Charlotte M CM

Journal of chemical information and modeling 20200804 8

Current deep learning methods for structure-based virtual screening take the structures of both the protein and the ligand as input but make little or no use of the protein structure when predicting ligand binding. Here, we show how a relatively simple method of data set augmentation forces such deep learning methods to take into account information from the protein. Models trained in this way are more generalizable (make better predictions on protein/ligand complexes from a different distributi ...[more]

PMID: 32701288

Similar Datasets

On TCR binding predictors failing to generalize to unseen peptides.

Project description:Several recent studies investigate TCR-peptide/-pMHC binding prediction using machine learning or deep learning approaches. Many of these methods achieve impressive results on test sets, which include peptide sequences that are also included in the training set. In this work, we investigate how state-of-the-art deep learning models for TCR-peptide/-pMHC binding prediction generalize to unseen peptides. We create a dataset including positive samples from IEDB, VDJdb, McPAS-TCR, and the MIRA set, as well as negative samples from both randomization and 10X Genomics assays. We name this collection of samples TChard. We propose the hard split, a simple heuristic for training/test split, which ensures that test samples exclusively present peptides that do not belong to the training set. We investigate the effect of different training/test splitting techniques on the models' test performance, as well as the effect of training and testing the models using mismatched negative samples generated randomly, in addition to the negative samples derived from assays. Our results show that modern deep learning methods fail to generalize to unseen peptides. We provide an explanation why this happens and verify our hypothesis on the TChard dataset. We then conclude that robust prediction of TCR recognition is still far for being solved.

| S-EPMC9634250 | biostudies-literature

Virtual data augmentation method for reaction prediction.

Project description:To improve the performance of data-driven reaction prediction models, we propose an intelligent strategy for predicting reaction products using available data and increasing the sample size using fake data augmentation. In this research, fake data sets were created and augmented with raw data for constructing virtual training models. Fake reaction datasets were created by replacing some functional groups, i.e., in the data analysis strategy, the fake data as compounds with modified functional groups to increase the amount of data for reaction prediction. This approach was tested on five different reactions, and the results show improvements over other relevant techniques with increased model predictivity. Furthermore, we evaluated this method in different models, confirming the generality of virtual data augmentation. In summary, virtual data augmentation can be used as an effective measure to solve the problem of insufficient data and significantly improve the performance of reaction prediction.

| S-EPMC9556613 | biostudies-literature

Myosin MyTH4-FERM structures highlight important principles of convergent evolution.

Project description:Myosins containing MyTH4-FERM (myosin tail homology 4-band 4.1, ezrin, radixin, moesin, or MF) domains in their tails are found in a wide range of phylogenetically divergent organisms, such as humans and the social amoeba Dictyostelium (Dd). Interestingly, evolutionarily distant MF myosins have similar roles in the extension of actin-filled membrane protrusions such as filopodia and bind to microtubules (MT), suggesting that the core functions of these MF myosins have been highly conserved over evolution. The structures of two DdMyo7 signature MF domains have been determined and comparison with mammalian MF structures reveals that characteristic features of MF domains are conserved. However, across millions of years of evolution conserved class-specific insertions are seen to alter the surfaces and the orientation of subdomains with respect to each other, likely resulting in new sites for binding partners. The MyTH4 domains of Myo10 and DdMyo7 bind to MT with micromolar affinity but, surprisingly, their MT binding sites are on opposite surfaces of the MyTH4 domain. The structural analysis in combination with comparison of diverse MF myosin sequences provides evidence that myosin tail domain features can be maintained without strict conservation of motifs. The results illustrate how tuning of existing features can give rise to new structures while preserving the general properties necessary for myosin tails. Thus, tinkering with the MF domain enables it to serve as a multifunctional platform for cooperative recruitment of various partners, allowing common properties such as autoinhibition of the motor and microtubule binding to arise through convergent evolution.

| S-EPMC4889382 | biostudies-literature

Predicting functionally important SNP classes based on negative selection.

Project description:BackgroundWith the advent of cost-effective genotyping technologies, genome-wide association studies allow researchers to examine hundreds of thousands of single nucleotide polymorphisms (SNPs) for association with human disease. Recently, many researchers applying this strategy have detected strong associations to disease with SNP markers that are either not in linkage disequilibrium with any nonsynonymous SNP or large distances from any annotated gene. In such cases, no well-established standard practice for effective SNP selection for follow-up studies exists. We aim to identify and prioritize groups of SNPs that are more likely to affect phenotypes in order to facilitate efficient SNP selection for follow-up studies.ResultsBased on the annotations available in the Ensembl database, we categorized SNPs in the human genome into classes related to regulatory attributes, such as epigenetic modifications and transcription factor binding sites, in addition to classes related to gene structure and cross-species conservation. Using the distribution of derived allele frequencies (DAF) within each class, we assessed the strength of natural selection for each class relative to the genome as a whole. We applied this DAF analysis to Perlegen resequenced SNPs genome-wide. Regulatory elements annotated by Ensembl such as specific histone methylation sites as well as classes defined by cross-species conservation showed negative selection in comparison to the genome as a whole.ConclusionsThese results highlight which annotated classes are under purifying selection, have putative functional importance, and contain SNPs that are strong candidates for follow-up studies after genome-wide association. Such SNP annotation may also be useful in interpreting results of whole-genome sequencing studies.

| S-EPMC3033802 | biostudies-literature

Better force fields start with better data: A data set of cation dipeptide interactions.

Project description:We present a data set from a first-principles study of amino-methylated and acetylated (capped) dipeptides of the 20 proteinogenic amino acids - including alternative possible side chain protonation states and their interactions with selected divalent cations (Ca2+, Mg2+ and Ba2+). The data covers 21,909 stationary points on the respective potential-energy surfaces in a wide relative energy range of up to 4 eV (390 kJ/mol). Relevant properties of interest, like partial charges, were derived for the conformers. The motivation was to provide a solid data basis for force field parameterization and further applications like machine learning or benchmarking. In particular the process of creating all this data on the same first-principles footing, i.e. density-functional theory calculations employing the generalized gradient approximation with a van der Waals correction, makes this data suitable for first principles data-driven force field development. To make the data accessible across domain borders and to machines, we formalized the metadata in an ontology.

| S-EPMC9205945 | biostudies-literature

I know better! Emerging metacognition allows adolescents to ignore false advice.

Project description:Adolescents aspire for independence. Successful independence means knowing when to rely on one's own knowledge and when to listen to others. A critical prerequisite thus is a well-developed metacognitive ability to accurately assess the quality of one's own knowledge. Little is known about whether the strive to become an independent decision maker in adolescence is underpinned by the necessary metacognitive skills. Here, we demonstrate that metacognition matures from childhood to adolescence (N = 107) and that this process coincides with greater independent decision-making. We show that adolescents, in contrast to children, take on others' advice less often, but only when the advice is misleading. Finally, we demonstrate that adolescents' reduced reliance on others' advice is explained by their increased metacognitive skills, suggesting that a developing ability to introspect may support independent decision-making in adolescence.

| S-EPMC8612133 | biostudies-literature

ASO Author Reflections: Advanced Imaging Allows Better Detection of Peritoneal Metastases.

Project description: Not available

| S-EPMC6338702 | biostudies-literature

Identification of new classes of ricin toxin inhibitors by virtual screening.

Project description:We used two virtual screening programs, ICM and GOLD, to dock nearly 50,000 compounds into each of two conformations of the target protein ricin A chain (RTA). A limited control set suggests that candidates scored highly by two programs may have a higher probability of being ligands than those in a list from a single program. Based on the virtual screens, we purchased 306 compounds that were subjected to a kinetic assay. Six compounds were found to give modest, but significant, inhibition of RTA. They also tended to inhibit Shiga toxin A chain, with roughly the same IC(50). The compounds generally represent novel chemical platforms that do not resemble RTA substrates, as currently known inhibitors do. These six were also tested in a cell-based assay for their ability to protect cells from intact ricin. Two compounds were effective in this regard, showing modest to strong ricin inhibition, but also showing some cytotoxicity. RTA, with its large, polar active site is a difficult drug design target which is expected to bind small molecules only weakly. The ability of the method to find these novel platforms is encouraging and suggests virtual screening can contribute to the search for ricin and Shiga toxin inhibitors.

| S-EPMC2929769 | biostudies-literature

New classes of potent heparanase inhibitors from ligand-based virtual screening.

Project description:Heparanase is a validated target in cancer therapy and a potential target for several inflammatory pathologies. A ligand-based virtual screening of commercial libraries was performed to expand the chemical space of small-molecule inhibitors. The screening was based on similarity with known inhibitors and was performed in several runs, starting from literature compounds and progressing through newly discovered inhibitors. Among the fifty-five tested compounds, nineteen had IC50 values lower than 5 µM and some showed remarkable potencies. Importantly, tere- and isophthalamides derivatives belong to new structural classes of heparanase inhibitors and some of them showed enzyme affinities (61 and 63, IC50 = 0.32 and 0.12 µM, respectively) similar to those of the most potent small-molecule inhibitors reported so far. Docking studies provided a comprehensive binding hypothesis shared by compounds with significant structural diversity. The most potent inhibitors reduced cell invasiveness and inhibited the expression of proangiogenic factors in tumour cell lines.

| S-EPMC7534336 | biostudies-literature

Data Augmentation for a Virtual-Sensor-Based Nitrogen and Phosphorus Monitoring.

Project description:To better control eutrophication, reliable and accurate information on phosphorus and nitrogen loading is desired. However, the high-frequency monitoring of these variables is economically impractical. This necessitates using virtual sensing to predict them by utilizing easily measurable variables as inputs. While the predictive performance of these data-driven, virtual-sensor models depends on the use of adequate training samples (in quality and quantity), the procurement and operational cost of nitrogen and phosphorus sensors make it impractical to acquire sufficient samples. For this reason, the variational autoencoder, which is one of the most prominent methods in generative models, was utilized in the present work for generating synthetic data. The generation capacity of the model was verified using water-quality data from two tributaries of the River Thames in the United Kingdom. Compared to the current state of the art, our novel data augmentation-including proper experimental settings or hyperparameter optimization-improved the root mean squared errors by 23-63%, with the most significant improvements observed when up to three predictors were used. In comparing the predictive algorithms' performances (in terms of the predictive accuracy and computational cost), k-nearest neighbors and extremely randomized trees were the best-performing algorithms on average.

| S-EPMC9920320 | biostudies-literature

OmicsDI is part of the ELIXIR infrastructure

OmicsDI is an Elixir interoperability service. Learn more ›

Tweets

OmicsDI Databases

PRIDE
PeptideAtlas
MassIVE
JPOST Repository
Physiome Model Repository

EGA
EVA
ENA
LINCS
PAXDB
Cell Collective

MetaboLights
Metabolomics Workbench
MetabolomeExpress
GNPS
BioModels
FAIRDOMHub

ArrayExpress
dbGaP
ExpressionAtlas
GEO
NODE

Information

Databases
Help
API
Contact us
Code on GitHub
Terms of use
Submit Data