Single-cell analyses have led to many breakthrough discoveries on the identification of celltypes and functional
states. However, downstream analyses aiming to infer causal gene regulatory networks driving cell differen-
tiation and cell-cell interactions remain challenging. More generally, causal networks are usually difficult to
learn and interpret regardless of the data type, as most algorithms (both constraint-based and score-based
approaches) are structurally stringent and data-restrictive. Our team has been developing a novel approach
called MIIC (Multivariate Information-based Inductive Causation) that efficiently combines constraint-based
and information-theoretic frameworks, which greatly improves the precision of inferred causal networks. Here
we showcase MIIC on single-cell transcriptomic data coming from an in-depth study of the induction of reg-
ulatory B cells called iBregs to cure central nervous system autoimmunity. We identified that B cells can
control autoimmunity through interleukin(IL)-10 production, and prevented the development of a multiple
sclerosis mice model called EAE as well as cured recipient mice within a few days upon administration of
the iBregs at the peak of clinical signs. However, the cellular interactions of the injected B cells causing
the mice to heal remain unknown. The inferred causal network of microglia samples using MIIC uncovered
gene regulation activities that will be tested in vitro, in an effort to reconstruct a causal model of the EAE
recovery in mice following iBregs injection.