Researchers in France are launching a new trial that will harness the power of artificial intelligence-based analysis and apply it to the identification of mechanisms underlying the response or resistance to immunotherapy with immune checkpoint inhibitors in patients with non-small cell lung cancer (NSCLC).1

In the first part of the study — the PIONEER trial — patients with NSCLC treated with immune checkpoint inhibitors will be carefully monitored with comprehensive searches for immune-related markers, gut microbiota, pharmacokinetics, pharmacogenomics, pharmacogenetics, and multiple biological parameters, explained study researcher Joseph Ciccolini, PharmD, PhD, professor of pharmacokinetics for the Center for Research on Cancer at Aix-Marseille University in France.  

“All of this knowledge will lead to an original data analysis step using sophisticated mathematical modeling — the QUANTIC part of the project — to help understand what makes patients respond (and survive) and what makes other patients progress on immunotherapy,” Dr Ciccolini said.

The goal of this study is an important one, said Lee A.D. Cooper, PhD, director of computational pathology and the center for computational imaging and signal analysis at Northwestern University Feinberg School of Medicine in Chicago, Illinois.


Continue Reading

“Understanding who will respond to these therapies is one of the fundamental challenges that exists right now,” Dr Cooper said. “Predicting response is a challenging problem. The QUANTIC team is not just looking at predicting who will respond to treatment, but [is] going beyond that — to understanding why these people respond.”

In their explanation of the combined PIONeeR/QUANTIC project, Pr. Ciccolini and colleagues explain that the data from PIONeeR will result in “hundreds of quantitative variables per time point per patient.”

To analyze the data, the researchers will use mechanistic modeling. This type of modeling is thought to have superior value to artificial intelligence because the data produced are interpretable.

“This allows [the models] to account for the biological meaning of part of the data, (eg, quantification of immune players), and to test biological hypotheses, which improves our mechanistic understanding of the processes at play,” they wrote.

Not all of the data will have biological meaning, though, so some part of the modeling will remain biologically agnostic and will rely on machine learning alone. Additionally, the researchers state that mixed-effect statistical learning will be used in the analysis.

“All patients’ data will be pooled together for the learning process, which strengthens estimation of the mechanistic parameters,” they wrote. “Machine learning for inclusion of baseline covariates will further yield new algorithms able to predict the response/relapse patterns, including possible pseudo- or hyperprogression.”