Identifying Differences in the Performance of Machine Learning Models for Off-Targets Trained on Publicly Available and Proprietary Data Sets.
Ontology highlight
ABSTRACT:
SUBMITTER: Smajic A
PROVIDER: S-EPMC10445286 | biostudies-literature | 2023 Aug
REPOSITORIES: biostudies-literature
Publications
Identifying Differences in the Performance of Machine Learning Models for Off-Targets Trained on Publicly Available and Proprietary Data Sets.
Smajić Aljoša A Rami Iris I Sosnin Sergey S Ecker Gerhard F GF
Chemical research in toxicology 20230713 8
Each year, publicly available databases are updated with new compounds from different research institutions. Positive experimental outcomes are more likely to be reported; therefore, they account for a considerable fraction of these entries. Established publicly available databases such as ChEMBL allow researchers to use information without constrictions and create predictive tools for a broad spectrum of applications in the field of toxicology. Therefore, we investigated the distribution of pos ...[more]
Similar Datasets
Project description:RNA viruses are abundant and highly diverse and infect all or most eukaryotic organisms. However, only a tiny fraction of the number and diversity of RNA virus species have been catalogued. To cost-effectively expand the diversity of known RNA virus sequences, we mined publicly available transcriptomic data sets. We developed 77 family-level Hidden Markov Model profiles for the viral RNA-dependent RNA polymerase (RdRp)-the only universal "hallmark" gene of RNA viruses. By using these to search the National Center for Biotechnology Information Transcriptome Shotgun Assembly database, we identified 5,867 contigs encoding RNA virus RdRps or fragments thereof and analyzed their diversity, taxonomic classification, phylogeny, and host associations. Our study expands the known diversity of RNA viruses, and the 77 curated RdRp Profile Hidden Markov Models provide a useful resource for the virus discovery community.
| S-EPMC10101049 | biostudies-literature
Project description:ImportanceThere are limited resources providing postdonation conditions that can occur in living donors (LDs) of solid-organ transplant. Consequently, it is difficult to visualize and understand possible postdonation outcomes in LDs.ObjectiveTo assemble an open access resource that is representative of the demographic characteristics in the US national registry, maintained by the Organ Procurement and Transplantation Network and administered by the United Network for Organ Sharing, but contains more follow-up information to help to examine postdonation outcomes in LDs.Design, setting, and participantsCohort study in which the data for the resource and analyses stemmed from the transplant data set derived from 27 clinical studies from the ImmPort database, which is an open access repository for clinical studies. The studies included data collected from 1963 to 2016. Data from the United Network for Organ Sharing Organ Procurement and Transplantation Network national registry collected from October 1987 to March 2016 were used to determine representativeness. Data analysis took place from June 2016 to May 2018. Data from 20 ImmPort clinical studies (including clinical trials and observational studies) were curated, and a cohort of 11 263 LDs was studied, excluding deceased donors, LDs with 95% or more missing data, and studies without a complete data dictionary. The harmonization process involved the extraction of common features from each clinical study based on categories that included demographic characteristics as well as predonation and postdonation data.Main outcomes and measuresThirty-six postdonation events were identified, represented, and analyzed via a trajectory network analysis.ResultsThe curated data contained 10 869 living kidney donors (median [interquartile range] age, 39 [31-48] years; 6175 [56.8%] women; and 9133 [86.6%] of European descent). A total of 9558 living kidney donors with postdonation data were analyzed. Overall, 1406 LDs (14.7%) had postdonation events. The 4 most common events were hypertension (806 [8.4%]), diabetes (190 [2.0%]), proteinuria (171 [1.8%]), and postoperative ileus (147 [1.5%]). Relatively few events (n = 269) occurred before the 2-year postdonation mark. Of the 1746 events that took place 2 years or more after donation, 1575 (90.2%) were nonsurgical; nonsurgical conditions tended to occur in the wide range of 2 to 40 years after donation (odds ratio, 38.3; 95% CI, 4.12-1956.9).Conclusions and relevanceMost events that occurred more than 2 years after donation were nonsurgical and could occur up to 40 years after donation. Findings support the construction of a national registry for long-term monitoring of LDs and confirm the value of secondary reanalysis of clinical studies.
| S-EPMC6481454 | biostudies-literature
Project description:The screening of compounds for ADME-Tox targets plays an important role in drug design. QSPR models can increase the speed of these specific tasks, although the performance of the models highly depends on several factors, such as the applied molecular descriptors. In this study, a detailed comparison of the most popular descriptor groups has been carried out for six main ADME-Tox classification targets: Ames mutagenicity, P-glycoprotein inhibition, hERG inhibition, hepatotoxicity, blood-brain-barrier permeability, and cytochrome P450 2C9 inhibition. The literature-based, medium-sized binary classification datasets (all above 1,000 molecules) were used for the model building by two common algorithms, XGBoost and the RPropMLP neural network. Five molecular representation sets were compared along with their joint applications: Morgan, Atompairs, and MACCS fingerprints, and the traditional 1D and 2D molecular descriptors, as well as 3D molecular descriptors, separately. The statistical evaluation of the model performances was based on 18 different performance parameters. Although all the developed models were close to the usual performance of QSPR models for each specific ADME-Tox target, the results clearly showed the superiority of the traditional 1D, 2D, and 3D descriptors in the case of the XGBoost algorithm. It is worth trying the classical tools in single model building because the use of 2D descriptors can produce even better models for almost every dataset than the combination of all the examined descriptor sets.
| S-EPMC9214226 | biostudies-literature
Project description:Malaria is a deadly disease caused by Apicomplexan parasites of the Plasmodium genus. Several species of the Plasmodium genus are known to be infectious to human, of which P. falciparum is the deadliest. Post-translational modifications (PTMs) of proteins coordinate cell signalling and hence, regulate many biological processes in P. falciparum homeostasis and host infection, of which the most highly studied is phosphorylation. Phosphosites on proteins can be identified by tandem mass spectrometry (MS) performed on enriched samples (phosphoproteomics), followed by downstream computational analyses. We have performed a large-scale meta-analysis of 11 publicly available phosphoproteomics datasets, to build a comprehensive atlas of phosphosites in the P. falciparum proteome, using robust pipelines aimed at strict control of false identifications. We identified a total of 28,495 phosphorylated sites on P. falciparum proteins at 5% false localisation rate (FLR) and, of those, 18,100 at 1% FLR. We identified significant sequence motifs, likely indicative of different groups of kinases, responsible for different groups of phosphosites. Conservation analysis identified clusters of phosphoproteins that are highly conserved, and others that are evolving faster within the Plasmodium genus, and implicated in different pathways. We also explored the structural context of phosphosites, identifying a strong enrichment for phosphosites on fast evolving (low conservation) intrinsically disordered regions (IDRs) of proteins. In other species, IDRs have been shown to have an important role in modulating protein-protein interactions, particularly in signalling, and thus warranting further study for their roles in host-pathogen interactions. All data has made available via UniProt, PRIDE and PeptideAtlas, with visualisation interfaces for exploring phosphosites in the context of other data on Plasmodium proteins.We have re-analysed publicly available mass spectrometry (MS) data sets enriched for phosphopeptides from Asian rice (Oryza sativa). In total we have identified, 15522 phosphosites on Serine, Threonine and Tyrosine residues on rice proteins. The data has been loaded into UniProtKB, enabling researchers to visualise the sites alongside other stored data on rice proteins, including structural models from AlphaFold2, and into PeptideAtlas, enabling visualisation of the source evidence for each site, including scores and source mass spectra. We identified sequence motifs for phosphosites, and link motifs to enrichment of different biological processes, indicating different downstream regulation caused by different kinase groups. We cross-referenced phosphosites against single amino acid variation (SAAV) data sourced from the rice 3000 genomes data, to identify SAAVs within or proximal to phosphosites that could cause loss of a particular site in a given rice variety. The data was further clustered to identify groups of sites with similar patterns across rice family groups, allowing us to identify sites highly conserved in Japonica, but mostly absent in, for example, Aus type rice varieties - known to have different responses to drought. These resources can assist rice researchers to discover alleles with significantly different functional effects across rice varieties.
2023-11-22 | PXD046874 | Pride
Project description:Many important questions remain regarding severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), the viral pathogen responsible for COVID-19. These questions include the mechanisms explaining the high percentage of asymptomatic but highly infectious individuals, the wide variability in disease susceptibility, and the mechanisms of long-lasting debilitating effects. Bioinformatic analysis of four coronavirus datasets representing previous outbreaks (SARS-CoV-1 and MERS-CoV), as well as SARS-CoV-2, revealed evidence of diverse host factors that appear to be coopted to facilitate virus-induced suppression of interferon-induced innate immunity, promotion of viral replication and subversion and/or evasion of antiviral immune surveillance. These host factors merit further study given their postulated roles in COVID-19-induced loss of smell and brain, heart, vascular, lung, liver, and gut dysfunction.
| S-EPMC8480897 | biostudies-literature
Project description:BackgroundThe COVID-19 pandemic has changed public health policies and human and community behaviors through lockdowns and mandates. Governments are rapidly evolving policies to increase hospital capacity and supply personal protective equipment and other equipment to mitigate disease spread in affected regions. Current models that predict COVID-19 case counts and spread are complex by nature and offer limited explainability and generalizability. This has highlighted the need for accurate and robust outbreak prediction models that balance model parsimony and performance.ObjectiveWe sought to leverage readily accessible data sets extracted from multiple states to train and evaluate a parsimonious predictive model capable of identifying county-level risk of COVID-19 outbreaks on a day-to-day basis.MethodsOur modeling approach leveraged the following data inputs: COVID-19 case counts per county per day and county populations. We developed an outbreak gold standard across California, Indiana, and Iowa. The model utilized a per capita running 7-day sum of the case counts per county per day and the mean cumulative case count to develop baseline values. The model was trained with data recorded between March 1 and August 31, 2020, and tested on data recorded between September 1 and October 31, 2020.ResultsThe model reported sensitivities of 81%, 92%, and 90% for California, Indiana, and Iowa, respectively. The precision in each state was above 85% while specificity and accuracy scores were generally >95%.ConclusionsOur parsimonious model provides a generalizable and simple alternative approach to outbreak prediction. This methodology can be applied to diverse regions to help state officials and hospitals with resource allocation and to guide risk management, community education, and mitigation strategies.
| S-EPMC8315156 | biostudies-literature
Project description:UnlabelledBackgroundIdentification of prognostic biomarkers is hallmark of cancer genomics. Since miRNAs regulate expression of multiple genes, they act as potent biomarkers in several cancers. Identification of miRNAs that are prognostically important has been done sporadically, but no resource is available till date that allows users to study prognostics of miRNAs of interest, utilizing the wealth of available data, in major cancer types.DescriptionIn this paper, we present a web based tool that allows users to study prognostic properties of miRNAs in several cancer types, using publicly available data. We have compiled data from Gene Expression Omnibus (GEO), and recently developed "The Cancer Genome Atlas (TCGA)", to create this tool. The tool is called "PROGmiR" and it is available at http://www.compbio.iupui.edu/progmir. Currently, our tool can be used to study overall survival implications for approximately 1050 human miRNAs in 16 major cancer types.ConclusionsWe believe this resource, as a hypothesis generation tool, will be helpful for researchers to link miRNA expression with cancer outcome and to design mechanistic studies. We studied performance of our tool using identified miRNA biomarkers from published studies. The prognostic plots created using our tool for specific miRNAs in specific cancer types corroborated with the findings in the studies.
| S-EPMC3564827 | biostudies-literature
Project description:BackgroundSince the introduction of microarrays in 1995, researchers world-wide have used both commercial and custom-designed microarrays for understanding differential expression of transcribed genes. Public databases such as ArrayExpress and the Gene Expression Omnibus (GEO) have made millions of samples readily available. One main drawback to microarray data analysis involves the selection of probes to represent a specific transcript of interest, particularly in light of the fact that transcript-specific knowledge (notably alternative splicing) is dynamic in nature.ResultsWe therefore developed a framework for reannotating and reassigning probe groups for Affymetrix® GeneChip® technology based on functional regions of interest. This framework addresses three issues of Affymetrix® GeneChip® data analyses: removing nonspecific probes, updating probe target mapping based on the latest genome knowledge and grouping probes into gene, transcript and region-based (UTR, individual exon, CDS) probe sets. Updated gene and transcript probe sets provide more specific analysis results based on current genomic and transcriptomic knowledge. The framework selects unique probes, aligns them to gene annotations and generates a custom Chip Description File (CDF). The analysis reveals only 87% of the Affymetrix® GeneChip® HG-U133 Plus 2 probes uniquely align to the current hg38 human assembly without mismatches. We also tested new mappings on the publicly available data series using rat and human data from GSE48611 and GSE72551 obtained from GEO, and illustrate that functional grouping allows for the subtle detection of regions of interest likely to have phenotypical consequences.ConclusionThrough reanalysis of the publicly available data series GSE48611 and GSE72551, we profiled the contribution of UTR and CDS regions to the gene expression levels globally. The comparison between region and gene based results indicated that the detected expressed genes by gene-based and region-based CDFs show high consistency and regions based results allows us to detection of changes in transcript formation.
| S-EPMC5731501 | biostudies-literature
Project description:ObjectiveAs data science and artificial intelligence continue to rapidly gain traction, the publication of freely available ICU datasets has become invaluable to propel data-driven clinical research. In this guide for clinicians and researchers, we aim to: 1) systematically search and identify all publicly available adult clinical ICU datasets, 2) compare their characteristics, data quality, and richness and critically appraise their strengths and weaknesses, and 3) provide researchers with suggestions, which datasets are appropriate for answering their clinical question.Data sourcesA systematic search was performed in Pubmed, ArXiv, MedRxiv, and BioRxiv.Study selectionWe selected all studies that reported on publicly available adult patient-level intensive care datasets.Data extractionA total of four publicly available, adult, critical care, patient-level databases were included (Amsterdam University Medical Center data base [AmsterdamUMCdb], eICU Collaborative Research Database eICU CRD], High time-resolution intensive care unit dataset [HiRID], and Medical Information Mart for Intensive Care-IV). Databases were compared using a priori defined categories, including demographics, patient characteristics, and data richness. The study protocol and search strategy were prospectively registered.Data synthesisFour ICU databases fulfilled all criteria for inclusion and were queried using SQL (PostgreSQL version 12; PostgreSQL Global Development Group) and analyzed using R (R Foundation for Statistical Computing, Vienna, Austria). The number of unique patient admissions varied between 23,106 (AmsterdamUMCdb) and 200,859 (eICU-CRD). Frequency of laboratory values and vital signs was highest in HiRID, for example, 5.2 (±3.4) lactate values per day and 29.7 (±10.2) systolic blood pressure values per hour. Treatment intensity varied with vasopressor and ventilatory support in 69.0% and 83.0% of patients in AmsterdamUMCdb versus 12.0% and 21.0% in eICU-CRD, respectively. ICU mortality ranged from 5.5% in eICU-CRD to 9.9% in AmsterdamUMCdb.ConclusionsWe identified four publicly available adult clinical ICU datasets. Sample size, severity of illness, treatment intensity, and frequency of reported parameters differ markedly between the databases. This should guide clinicians and researchers which databases to best answer their clinical questions.
| S-EPMC9150442 | biostudies-literature
Project description:BACKGROUND: The criteria for choosing relevant cell lines among a vast panel of available intestinal-derived lines exhibiting a wide range of functional properties are still ill-defined. The objective of this study was, therefore, to establish objective criteria for choosing relevant cell lines to assess their appropriateness as tumor models as well as for drug absorption studies. RESULTS: We made use of publicly available expression signatures and cell based functional assays to delineate differences between various intestinal colon carcinoma cell lines and normal intestinal epithelium. We have compared a panel of intestinal cell lines with patient-derived normal and tumor epithelium and classified them according to traits relating to oncogenic pathway activity, epithelial-mesenchymal transition (EMT) and stemness, migratory properties, proliferative activity, transporter expression profiles and chemosensitivity. For example, SW480 represent an EMT-high, migratory phenotype and scored highest in terms of signatures associated to worse overall survival and higher risk of recurrence based on patient derived databases. On the other hand, differentiated HT29 and T84 cells showed gene expression patterns closest to tumor bulk derived cells. Regarding drug absorption, we confirmed that differentiated Caco-2 cells are the model of choice for active uptake studies in the small intestine. Regarding chemosensitivity we were unable to confirm a recently proposed association of chemo-resistance with EMT traits. However, a novel signature was identified through mining of NCI60 GI50 values that allowed to rank the panel of intestinal cell lines according to their drug responsiveness to commonly used chemotherapeutics. CONCLUSIONS: This study presents a straightforward strategy to exploit publicly available gene expression data to guide the choice of cell-based models. While this approach does not overcome the major limitations of such models, introducing a rank order of selected features may allow selecting model cell lines that are more adapted and pertinent to the addressed biological question.
| S-EPMC3412164 | biostudies-literature