The doctoral dissertation in the field of Computational Biomedicine will be examined at the Faculty of Health Sciences at Kuopio campus. The public examination will be streamed online.
What is the topic of your doctoral research? Why is it important to study the topic?
My doctoral research concerns the application of methods based on artificial intelligence (AI) and machine learning for grouping patients based on their omics profiles. Grouping or stratifying patients can be used to identify novel molecular subtypes of diseases. Furthermore, by integrating prior biological knowledge, including functional annotations and gene interactions, with omics profiles, we can address noise factors more effectively. We developed and applied novel knowledge integration methods to discover omics-based subtype in several common cancer types. Omics-profiles exhibit significant heterogeneity between cancer patients compounded by biological and technical noise factors. Therefore, dividing cancers into subtypes helps identify important characteristics and molecular mechanisms more precisely. A key goal of the precision medicine paradigm is to develop targeted therapies that can be used to better tailor treatment plans based on individual patient needs. Such treatments have the potential to improve efficacy while avoiding side effects. Due to the size and complexity of omics data, computational methods are essential for studying precision medicine. Advanced AI methods can integrate different data types and background knowledge more effectively than traditional methods, making them increasingly important in biomedical research.
What are the key findings or observations of your doctoral research?
Integrating prior biological knowledge within clustering algorithms improves the clinical significance of patient stratification results by identifying groups with greater survival differences. Utilising prior knowledge in driving clustering analysis has been neglected in previous studies comparing different approaches. Furthermore, the evaluation methods of previous omics-based clustering benchmark studies could be improved. Hence, new software for evaluating and comparing approaches based on clustering quality, significance, and reproducibility was developed and utilised in new benchmark studies. The performances of algorithms were highly dependent on the dataset suggesting that testing multiple algorithms and performing comprehensive multi-criteria evaluation is crucial for identifying the best subtype candidates for further analysis.
How can the results of your doctoral research be utilised in practice?
My doctoral research has led to the development new methods, published as open-source software, which can be used to analyse new disease datasets. These methods are designed to advance the discovery of disease subtypes and facilitate targeted drug development. Computational biomedicine utilises existing datasets to generate and prioritise new hypotheses. This approach can lead to more individualised and effective treatment of various diseases over the long term. The latest AI model developed during my research can estimate drug sensitivity on individual patients. This capability could be used to identify drug repositioning candidates. However, the validation of the model is still in progress and additional studies are required to confirm its practical applicability.
What are the key research methods and materials used in your doctoral research?
Several methods based on artificial intelligence and machine learning were utilised in the research, including multi-task deep neural networks and pathway-based multi-omic integrating kernel-methods. The methods developed for biological knowledge integration used network algorithms such as random walks and node centrality as well as gene network analysis and pathway enrichment methods to define pathway-level expression profiles which were then used in patient stratification. The computational work was performed using supercomputers of the UEF Bioinformatics Center and the CSC. The primary patient dataset used in the study was The Cancer Genome Atlas (TCGA) project. TCGA provides public access to multiple different omics profiles from around 10,000 patient tumours for over 30 cancer types. The thesis study focused on several different cancer types, especially breast and prostate cancers, because they are relatively common. In addition to patient data, the study utilised public cancer cell-line omics-profiles from the Cancer Cell Line Encyclopedia (CCLE) and public high-throughput drug-screening data from the Cancer Therapeutics Response Portal (CTRP). In total the number of cell-lines was over 1,000 and the number of cell-line-drug-pairs with sensitivity data was over 300,000.
The doctoral dissertation of Teemu Rintala, MSc (Tech), entitled Artificial intelligence for omics-driven patient stratification and precision medicine will be examined at the Faculty of Health Sciences. The Opponent in the public examination will be Professor Sampsa Hautaniemi of the University of Helsinki, and the Custos will be Associate Professor Vittorio Fortino of the 91天美. The public examination will be held in English.
