Speaker
Description
Understanding causal mechanisms in biological systems is essential for decoding complex physiological and pathological processes. Causal learning, a branch of machine learning, establishes quantitative relationships between molecular variables, revealing regulatory dynamics. Identifying interactions between master regulators, their target transcripts, and external interventions is crucial for uncovering actionable targets in gene regulation.
Using single-cell RNA sequencing (scRNA-seq), we infer gene regulatory networks (GRNs) through classical statistical and causal learning approaches. While traditional correlation-based techniques capture co-expression patterns, they fail to distinguish direct regulatory interactions and causal directionality. In contrast, causal learning methods, such as probabilistic graphical models, directed acyclic graphs (DAGs), and Bayesian networks, enable a mechanistic understanding of transcriptional regulation by differentiating direct from indirect effects and identifying key causal regulators.
Once a GRN is established, we investigate the interplay between transcription factors (TFs), exogenous signalling, and endogenous gene products. Causal inference methodologies help deconvolute complex regulatory hierarchies and predict perturbation effects on cellular states. To further enhance causal discovery, we integrate causal kinetic models, which use dynamical systems approaches to infer gene regulatory interactions from pseudo-time-resolved scRNA-seq data. These models reconstruct transcriptional dynamics by capturing temporal dependencies and quantitative relationships in gene expression patterns, improving the identification of causal drivers of cellular state transitions and actionable targets.
A critical application of this approach is in understanding the molecular mechanisms underlying NeuroCOVID, and the neurological complications associated with SARS-CoV-2 infection. By leveraging causal learning on sc-multi-omic data from affected neural and microglial cells, we can unravel the acute from long-lasting dysregulated programs, identify key transcription factors driving chronic neuroinflammation, and map gene regulatory changes linked to neurological dysfunction.
This approach improves the interpretability of inferred GRNs and lays the foundation for novel therapeutic strategies in precision medicine. Such insights might facilitate the identification of therapeutic targets for mitigating LongCOVID-related pathologies.
| Author(s) | Vittorio Aiello*, Nicolò Caporale, Emanuele Villa, Giuseppe Testa |
|---|---|
| Affiliation(s) | HT/UNIMI, HT/UNIMI, HT, HT/UNIMI |
