x
[contact-form-7 404 "Not Found"]

ESMO 2023: What to Take Away. Several breakthroughs in advanced cancer detection.

Non-small cell lung cancer: Artificial Intelligence enables the preliminary identification of complementary survival signatures that involve previous therapies

A new method has been developed by he Centre Léon Bérard in Lyon (France) to improve the prognosis of on-small cell lung cancer (NSCLC), one of the most common cancers in the world. The Centre Léon Bérard has developed a HOT score which, based on gene expression, predicts patient survival. This process was carried out with the help of the KEM® (Knowledge Extraction and Management) explanatory artificial intelligence platform, which systematically extracts association rules between variables in a database and refines the HOT score.

Using this new tool, the analysis identified 4 genes that predict survival.  The results have yet to be confirmed by the Cancer Research Institute’s iAtlas, a database containing gene expression for more than 1,100 cancer patients across five different tissue types. Specific adjustments may still be required, but this is a strong advance made with the help of the KEM® explanatory artificial intelligence platform.

Poster
Abstract

Identification of biomarkers of survival across multiple cancer types using eXplainable Artificial Intelligence

The study conducted here aims to demonstrate the biological relevance of OncoKEM®, an AI-based therapeutic recommendation tool that calculates scores for up to 205 drugs based on the drug’s transcriptional signatures and the tumor’s transcriptional profile. To achieve this, the study seeks to compare the results of OncoKEM® with those obtained through a standard analysis of biological pathways. The explainable Artificial Intelligence (xAI) platform, KEM® (Knowledge Extraction and Management), was used to identify biomarkers characterizing patients with higher chances of survival based on a list of biomarkers, including drug scores generated by OncoKEM®.

The method employed is as follows: Data was derived from the PROFILER study, a molecular screening program. It was aggregated into a database with 247 patients and 215,670 variables, including information on survival, baseline characteristics, gene expression, REACTOME pathway dysregulations, and OncoKEM® scores. The KEM® xAI platform identified 55,335 relations linking candidate biomarkers to survival. These results were filtered based on support, lift, and statistical significance. The remaining relations were separated into two sets to study the associations between survival and, respectively, pathway dysregulations or OncoKEM® scores.

This study has, therefore, demonstrated the consistency and biological relevance of OncoKEM® and paves the way for using this tool as a prognostic marker for refractory cancers.

Poster
Abstract