Dissecting cell identity via network inference and in silico gene perturbation

Area of Science:

• Developmental Biology
• Systems Biology
• Computational Biology

Background:

• Cell identity is determined by intricate gene expression regulation, forming gene-regulatory networks.
• Understanding these networks is crucial for deciphering developmental processes and cell differentiation.

Purpose of the Study:

• To develop and validate CellOracle, a machine learning approach for simulating transcription factor perturbations in gene regulatory networks.
• To analyze cell identity regulation and gain mechanistic insights into development and differentiation.

Main Methods:

• Inferred gene-regulatory networks from single-cell multi-omics data.
• Performed in silico transcription factor perturbations using CellOracle.
• Applied the approach to mouse/human hematopoiesis and zebrafish embryogenesis.

Main Results:

• CellOracle accurately modeled known phenotypic changes from transcription factor perturbations in established models.
• Simulated and experimentally validated a novel zebrafish phenotype resulting from the loss of the notochord regulator 'noto'.
• Identified 'lhx1a' as a novel axial mesoderm regulator in zebrafish.

Conclusions:

• CellOracle is a powerful tool for analyzing transcription factor roles in cell identity regulation.
• The approach provides valuable mechanistic insights into developmental processes and cell differentiation.
• CellOracle facilitates the discovery of novel gene functions and developmental phenotypes.