Inferring cascade drivers of VEXAS syndrome by a causal machine learning tool CauNagi

VEXAS syndrome, a severe autoinflammatory disease affecting adults, is caused by somatic mutations in the UBA1 gene. The disease's progression from primitive hematopoietic abnormalities to mature myeloid dysfunctions remains poorly understood. To tackle this challenge, researchers developed CauNagi, a computational framework designed to prioritize cascade candidate regulators (CCRs).

CauNagi combines a causal representation learning module from CausCell with an iterative deep learning backbone from UNAGI, and adds a unique downstream module for CCRs analysis.

The framework iteratively integrates causal disentangled representation learning with disease-stage cell-state trajectory reconstruction and dynamic regulatory analysis. Benchmarked on single-cell transcriptomic datasets, CauNagi successfully maintained cell-type structure in idiopathic pulmonary fibrosis and highlighted known AML-associated genes among its top-ranked global regulators.

When applied to VEXAS syndrome, CauNagi uncovered inflammatory responses, endoplasmic reticulum stress, and myeloid bias, aligning with the disease's features.

The CCRs analysis module of CauNagi identified 36 causal drivers, with SPI1, NFKB1, STAT3, and FOS emerging as high-confidence regulatory hubs. These findings were further validated using an independent single-cell transcriptomic dataset from a murine VEXAS model. CauNagi's computational efficiency and systematic approach make it a promising tool for identifying candidate causal regulators in other progressive disorders with available multistage single-cell transcriptomic datasets. CauNagi is accessible at https://github.com/steamed-stuffed-bun/CauNagi.