Insight

Eytan Ruppin: We present IRIS: a machine-learning method aimed at identifying candidate ligand-receptor interactions that are likely to mediate immune checkpoint blockade resistance in the tumor microenvironment.

Photo taken from Eytan Ruppin/X

Sep 24, 2023, 18:21

Eytan Ruppin: We present IRIS: a machine-learning method aimed at identifying candidate ligand-receptor interactions that are likely to mediate immune checkpoint blockade resistance in the tumor microenvironment.

Quoting Eytan Ruppin, Chief at the Cancer Data Science Lab of the National Institute of Health (NIH), on X/Twitter:

“Happy to share our recent work out on bioRxiv. We present IRIS: a machine-learning method aimed at identifying candidate ligand-receptor interactions that are likely to mediate immune checkpoint blockade (ICB) resistance in the tumor microenvironment.

IRIS takes as input: 1) ligand-receptor interaction activity profiles for each tumor sample, inferred from bulk transcriptomics harnessing our previously published tools CODEFACS and LIRICS and 2) the corresponding ICB response label for each sample.

IRIS then employs a two-step supervised machine learning method to identify a set of resistance activated (RAIs) and resistance deactivated interactions (RDIs), which are inferred to lead to ICB resistance. We applied IRIS to the 5 largest melanoma bulk RNA-seq ICB cohorts.

Strikingly, we found that the activity of deactivated interactions (RDI/RDS) offer markedly stronger predictive value for ICB therapy response compared to activated ones.

Moreover, our inferred RDIs (RDS) predictive accuracy is superior to that of state-of-the-art published transcriptomics biomarkers across an array of melanoma bulk RNA-seq ICB cohorts.

Focusing on resistance deactivated interactions, we find that many of these interactions are involved in stimulatory chemokine signaling, and that their activity is correlated with both CD8+ T cell infiltration levels and enriched in “hot” (brisk) tumor niches in TCGA melanoma.

Taken together, this study proposes a new strongly predictive immunotherapy response biomarker, showing that following ICB treatment, resistant tumors inhibit lymphocyte infiltration by deactivating specific key ligand-receptor interactions.