An information-theoretic framework for inferring T cell dynamics from single-cell RNA and T cell receptor sequencing.

TCRi infers joint clonotype–phenotype distributions in 167,118 CD8 T cells from cancer patients on anti-PD-(L)1 therapy, where mutual information robustly discriminates therapeutic response.

• Why it matters: Joint probabilistic modeling of clonotype and transcriptional phenotype from paired single-cell data has been limited, leaving uncertainty about how to quantify linked clonal dynamics and phenotypic diversity.
• What they did: TCRi uses a variational autoencoder to embed single-cell expression into a structured latent space, then applies a hierarchical Bayesian model to learn clonotype-specific phenotype distributions with information sharing across clonotypes, patients, and covariates.
• The result: From inferred distributions, it computes clonotypic entropy, phenotypic entropy, phenotypic flux, and mutual information; across four datasets, mutual information is a robust discriminator of therapeutic response.

AI summary of the abstract · may contain errors · not medical advice · read the paper.