Claudia Skok GibbsResearch Analyst, Systems Biology, CCB, Flatiron Institute
Presenters: Claudia Skok Gibbs, Research Analyst, Systems Biology
Chris Jackson, Postdoctoral Research Associate, New York University
Topic: Gene Regulatory Network Inference with the Inferelator
Understanding how gene expression programs are controlled requires identifying regulatory relationships between transcription factors and target genes. We have been actively developing the Inferelator as a python package for gene regulatory network inference that is based on regularized regression; this package takes gene expression data and a known network of prior interactions to learn a new network. Genome-wide expression data can be generated from microarrays or RNA-seq; to date, more than three million expression experiments have been loaded into NCBI GEO. Single-cell RNA sequencing (scRNAseq) has become increasingly popular for its ability to capture the gene expression state of thousands of individual cells in a single experiment. This offers advantages in combinatorial experimental design, large numbers of independent measurements, and accessing the interaction between the cell cycle and environmental responses that is hidden by population-level analysis of gene expression. The largest individual scRNAseq datasets have millions of individual samples. Technical limitations include each individual cell data point to be undersampled, and single-cell expression data is noisy and prone to batch effects. We have developed a technique for microfluidic scRNAseq in the budding yeast Saccharomyces cerevisiae, produced a data set of 40k cells that contain genetic and environmental perturbation, and used that data to infer a gene regulatory network using a multi-task implementation of the Inferelator. We demonstrate that the advantages of single-cell experimental design outweigh the technical disadvantages for regulatory network inference.