Description
Many biomedical studies profile patients using multiple modalities (e.g. various omics and imaging technologies), but there is a lack of knowledge-driven approaches of how to best use these multimodal datasets together to improve patient management. Further, most real-world healthcare datasets contain high degrees of missing data, together with high-dimensional and correlated genomic and molecular features, which hinder the use of standard ML approaches in treatment response modelling and prediction. This project will develop new knowledge-driven ML methods to make the most of these incomplete datasets for personalized treatment recommendations using logic-based decision rules and decision systems, where biological and clinical knowledge guides the search space exploration. We will develop ML approaches that allow data integration even in the presence of data missing not at random, enabling prediction without imputation but by exploiting the modalities redundancy structure. Therefore, a key theme in this project is redundancy in the heterogeneous data sources.
Specific project requirements
-
Master degree in machine learning, theoretical computer science, data science, (bio)statistics, bioinformatics, mathematics, or other relevant field
-
Good programming skills (e.g. Python, R) and ability to work with version control tools (e.g. Git)
-
Experience in bioinformatics data analysis tools for large-scale omics data is an advantage but is not necessary, as it can be learned
- Contract beginning no later than 1 October 2024
- This PhD position has four instead than three years, because 25% of the time will be dedicated to teaching and/or administrative tasks.
Supervisors
-
Associate Professor Valeria Vitelli, valeria.vitelli@medisin.uio.no contact person for inquiries about the position