During my time in Aix-Marseilles, I met with Prof Dominique Barbolosi and we started working together on a very different topic, that had attracted me for a long time: mixing our knowledge in mathematics and in computer science, to apply it to practical problems in medicine.

He introduced me to colleagues on the medical side such as Dr Xavier Muracciole.

Our first work together was on the MARLEN cohort, a longitudinal study of pediatric patients. After a medulloblastoma, treated by surgery, radiotherapy, and chemotherapy, the patients of that cohort were assessed every few years so as to monitor the development of their cognitive abilities.

We used various methods from statistics, Bayesian learning/Markov chains and machine learning to investigate the effect of covariates such as age, gender, or the presence of hydrocephalus, on the subsequent cognitive impairment and its progression over the years.

Publications: 55th Congress of the International Society of Paediatric Oncology (2023), full version close to submission (authors: X Muracciole, C Dufour, V Kieffer, A.Longaud, S Bolle, V Martin, C Grellois, N Andre, L Padovani, D Barbolosi)

Among other projects, we studied a cohort of patients treated for a melanoma from AP-HM (Marseilles) so as to predict the likelihood of a relapse. This matters, as doctors can prescribe prophylactic treatments, but these have side effects for the patient and also a rather significant cost for the French health system.

We used the Monolix software to:

• describe the tumoral growth via a simplified ODE model, with patient-dependent parameters,

• obtain the best fit for these parameters (via Monolix, which does a cohort-level computation)

• use the inferred individual parameters for (probabilistic) classification via more traditional machine learning approaches.

This approach combines mechanistic learning for precision medicine with more traditional machine learning techniques.

Publication: full version close to submission (authors: D. Barbolosi, C. Grellois, M. Canalis-Durand, C. Gaudy-Marqueste)