Big data to predict response to immunotherapy

Only 20% of bladder cancer patients respond to immunotherapy treatment. Now, research teams at PRBB have used massive data analysis to identify key markers of bladder cancer that may help determine the level of response to immunotherapy. The team, coordinated by Mar Albà (co-director of the Research Programme on Biomedical Informatics (GRIB)), Júlia Perera and Joaquim Bellmunt (Hospital del Mar Research Institute) and with the collaboration of Robert Castelo (Department of Medicine and Life Sciences, MELIS-UPF) and others, has analysed data from more than 700 bladder cancer patients from six different cohorts.

The researchers have developed a machine learning algorithm to see how patients respond to immunotherapy according to bladder cancer subtype. Using the algorithm, they have found that the neuronal subtype, although the least common of the five subtypes, is the one that responds best to treatment.

Thanks to the algorithm, they found that the markers that best determine treatment success are the number of mutations introduced by the APOBEC enzyme family, the number of mutations in the tumour and the number of anti-inflammatory macrophages. They have also identified some markers in the tumour microenvironment and, thanks to the large number of patients, have detected new rare mutations that would make the tumour invisible to the immune system.

“It is important to understand the mechanisms of response within the different subgroups, because there is a great deal of tumour heterogeneity”

Lilian Marie Boll, first author of the paper

This study helps to analyse the response mechanisms of different bladder cancer subtypes to immunotherapy. It also highlights the great diversity of tumours, which complicates their treatment. The research team stresses the need for a large amount of data in order to develop predictive models to treat individual tumours.