Página 1 dos resultados de 6430 itens digitais encontrados em 0.017 segundos
Resultados filtrados por Publicador: Universität Tübingen

‣ Interpretable Machine Learning Approaches in Computational Biology; Interpretierbare Maschinelle Lernansätze in der Bioinformatik

Briesemeister, Sebastian
Fonte: Universität Tübingen Publicador: Universität Tübingen
Tipo: Dissertation; info:eu-repo/semantics/doctoralThesis
Português
Relevância na Pesquisa
56.949146%
Machine learning has become an essential tool for analyzing, predicting, and understanding biological properties and processes. Machine learning models can substantially support the work of biologists by reducing the number of expensive and time-consuming experiments. They are able to uncover novel properties of biological systems and can be used to guide experiments. Machine learning models have been successfully applied to various tasks ranging from gene prediction to three-dimensional structure prediction of proteins. However, due to their lack of interpretability, many biologists put only little trust in the predictions made by computational models. In this thesis, we show how to overcome the typical "black box" character of machine learning algorithms by presenting two novel interpretable approaches for classification and regression. In the first part, we introduce YLoc, an interpretable classification approach for predicting the subcellular localization of proteins. YLoc is able to explain why a prediction was made by identifying the biological properties with the strongest influence on the prediction. We show that interpretable predictions made by YLoc help to understand a protein's localization and, moreover, can assist biologists in engineering the location of proteins. Furthermore...