Chemical representation learning for toxicity prediction
Born J, Markert G, Janakarajan N, Kimber T, Volkamer A, Martínez M, Manica M. Chemical representation learning for toxicity prediction. Digital Discovery 2023, 2: 674-691. DOI: 10.1039/d2dd00099g.Peer-Reviewed Original ResearchChemical language modelsLanguage modelMolecular property prediction tasksMolecular property prediction modelProperty prediction tasksMolecular property predictionExplicit supervisionAttention weightsMultiscale convolutionData augmentationPrediction taskToxicity datasetMolecular representationsProperty prediction modelsImproved accuracyModel reliabilityDatasetProperty predictionChemical representationsToxicity predictionPrediction uncertaintyUncertainty estimationDrug discoveryRepresentationPrediction modelFLAN: feature-wise latent additive neural models for biological applications
Nguyen A, Vasilaki S, Martínez M. FLAN: feature-wise latent additive neural models for biological applications. Briefings In Bioinformatics 2023, 24: bbad056. PMID: 37031956, PMCID: PMC10199769, DOI: 10.1093/bib/bbad056.Peer-Reviewed Original ResearchConceptsLearning modelsDeep neural networksDeep learning modelsMachine learning modelsBenchmark datasetsLatent spaceNeural modelNeural networkAlgorithmic decisionsEnd-usersBlack-boxComplex datasetsAggregated representationDecision processCritical scenariosImpressive resultsDatasetIndividual featuresRepresentationBiological domainData availabilityCodeUsersDecisionPerformanceIs Attention Interpretation? A Quantitative Assessment on Sets
Haab J, Deutschmann N, Martínez M. Is Attention Interpretation? A Quantitative Assessment on Sets. Communications In Computer And Information Science 2023, 1752: 303-321. DOI: 10.1007/978-3-031-23618-1_21.Peer-Reviewed Original ResearchBinary classification problemInterpretation of attentionClassification problemAttention mechanismSynthetic datasetsUnordered collectionClassification performanceSilent failuresMachine learningGlobal labelsData modalitiesIndividual instancesAttention distributionAttention scoresAttention patternsData pointsSub-componentsInstancesDatasetMonoNet: enhancing interpretability in neural networks via monotonic features
Nguyen A, Moreno D, Le-Bel N, Martínez M. MonoNet: enhancing interpretability in neural networks via monotonic features. Bioinformatics Advances 2023, 3: vbad016. PMID: 37143924, PMCID: PMC10152389, DOI: 10.1093/bioadv/vbad016.Peer-Reviewed Original ResearchNeural networkMonotonicity constraintsHigh-stakes scenariosInformation-theoretic analysisMachine learning modelsMedical informaticsNeural modelLearning capabilityLearning modelsBioinformatics Advances</i>Monotonous featuresComputational biologyEnhance interpretationModeling capabilitiesDatasetInterpretable modelsLearning processSample dataNetworkPower modelLearningSupplementary dataConstraintsPerformanceInformatics