Meno:František Václav
Priezvisko:Man
Názov:Automated Classification of the Clinical Impact of Structural Genomic Variations
Vedúci:Mgr. Jaroslav Budiš, PhD.
Rok:2026
Kľúčové slová:Multiple-instance learning, MIL, miGraph, MILES, structural variants, SV, copy number variants, CNV
Abstrakt:Modern biomedical technologies have enabled the detection of structural variations in a human genome. Assessing whether the detected structural variation has clinical effect has however remained a complex task, even for experienced laboratory diagnosticians. Existing computational tools often oversimplify input data used to better suite traditional machine learning methods, representing each variant as a single vector of data. As the number of affected genes differs between the variants, more flexible modelling approaches are necessary to avoid losing detailed gene-specific information. Our approach on prediction of clinical effect of CNV is based on two improvements. Firstly, we use comprehensive gene annotations to enrich the feature space. . Secondly, we shift from the traditional machine learning approaches to multiple instance learning that allows to model input data as a bags of feature instances. Such approach allows to use all gene information to better account for the varying pathogenic contributions of multiple genes within a single structural variation.

Súbory diplomovej práce:
Autor nedal súhlas so zverejnením svojej diplomovej práce.

Súbory prezentácie na obhajobe:

Upraviť