Mechanistic gene networks inferred from single-cell data with an outlier-insensitive method

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Single-cell techniques enable gene dynamics measurement at cellular resolution.
• Increased data complexity poses computational challenges for understanding biological mechanisms.
• Novel computational methods are crucial for predictive insights from complex single-cell data.

Purpose of the Study:

• To develop and present an iterative regression algorithm for inferring mechanistic gene networks from single-cell data.
• To address challenges posed by measurement outliers in single-cell gene expression data.
• To infer a developmental model for gene dynamics in Drosophila melanogaster blastoderm embryos.

Main Methods:

• An iterative regression algorithm was developed for gene network inference.
• The algorithm was applied to single-cell data from Drosophila melanogaster embryos.
• Model performance was compared against least squares, ridge regression, and neural networks.

Main Results:

• The inferred mechanistic model demonstrated superior predictive power compared to least squares and ridge regressions.
• Model predictions showed higher accuracy than various neural network architectures, even with small sample sizes.
• Predictions for gene knockouts showed substantial qualitative agreement with published experimental results.

Conclusions:

• The developed iterative regression algorithm effectively infers mechanistic gene networks from complex single-cell data.
• The inferred models provide accurate predictions of gene dynamics and developmental processes.
• This approach offers a powerful tool for understanding gene regulation and predicting responses to perturbations.