Bioinformatics and Big Data

Research focus

The Jonassen group works on development and application of bioinformatics methods contributing to the understanding of tumors and their microenvironments, aiming to aid in selecting appropriate treatments and predict outcome. They are currently working on a system medicine approach utilizing machine learning approaches targeting leukemia and development of methods to exploit the Hyperion technology to the study of tumormicroenvironment interactions in solid cancers. 

Subprojects

Jonassen leads the project AML_PM funded by ERAPerMed, including Bjørn Tore Gjertsen as a partner from CCBIO in addition to groups from Germany, the Netherlands and Canada. A postdoc in Jonassen’s group is working on developing and applying methods for analysis of various omics and single cell data generated by the partners. In this project, the group applies systems biology modeling and machine learning approaches aimed at predicting outcome and aid selection of treatment for individual patients, using a set of different experimental model systems and piloting clinical trials. For example, in a collaborative project with the Gjertsen group, results are promising, identifying single cell markers correlated with leukemia patients’ treatment response and survival. Another postdoc associated with CCBIO is working on development and use of methods to exploit the Hyperion imaging technology to the study of tumor microenvironment interactions. Pipelines including identification and annotation of individual cells have been established and current work includes analyzing a data set generated in the Akslen group encompassing a sample of breast tumors with associated outcome data. Jonassen expects a number of publications to result from the work in the coming year.

Recent important results

Relevant to Jonassen’s work in CCBIO, he published (in BMC Bioinformatics) in 2020 a flexible and versatile workflow for RNAseq data analysis, (in BMC Genomics) a comprehensive study comparing alternative approaches for characterizing DNA copy number variants, and (in Acta Neuropathologica Communications) a study showing that expression signatures seen in Parkinson are mainly driven by cell type composition. The latter work has relevance in analysis of leukemia and solid tumor data analysis where the group is now using single cell data to be able to better dissect changes in gene expression and relations to cell types and tumor microenvironments.

Current challenges

The Jonassen group aims to develop and use mathematical models that capture and predict effects of drugs targeting signaling molecules. Through the AML_PM project, they have established collaborations with groups having a strong track record in this area. In order to use such models to aid in selecting therapies for individual patients, they aim to utilize machine learning methods. One challenge is the relatively small size of training data that will be available for such approaches. The group’s approach will be to summarize the data and model predictions using a small number of parameters enabling learning from smaller training sets. A more technical challenge is the increasing focus from research funding agencies on data management plans and FAIR data sharing. This requires bioinformatics support, but also systematic efforts from those collecting samples and generating data in order to capture and describe in standardized ways meta-data allowing data reuse.

Future plans

Presently, the group is seeing promising results analyzing both the CYTOF and the Hyperion data. More work is needed to be able to fully exploit the data and to use it together with other data modalities on the same patients. To support this work, Jonassen and collaborators are currently preparing a new EraPerMed application following up the work started in the AML_PM project. The group will explore if some of the work, including that started on analysis of Hyperion data, can in part be linked with new institutional and medical faculty initiatives on artificial intelligence.