Benchmarking the prediction of responding cells to perturbations affecting both gene expression and cellular abundance using scRNA sequencing

Defining how cells respond to perturbations is crucial for understanding their biology. While previous works were designed to classify responding cells, methods for quantifying the response of each cell are not well developed. In addition, perturbations affecting only gene expression were considered for modeling. However, perturbation not only alters the gene expression of cells but also alters cellular abundance. But to date, no validation data is available to assess such perturbations. To address this issue, we utilized the clonal expansion of T or B cells as a new validation data type that can accompany both gene expression and alterations in cellular abundance after perturbation. Subsequently, we developed BRiCE, a new benchmark pipeline for assessing RC and Res, where RC refers to how well it can distinguish the responding cells against non-responding cells, while Res refers to how well it can quantify the responsiveness, and compares its performance with that of preexisting methods. However, none of the existing methods demonstrated predictive power. These results indicated that the current approach to understanding perturbation solely by gene expression is insufficient, and cellular abundance must be considered in perturbation modeling.