Using RNAi imaging data for gene interaction network construction

摘要

Embodiments of the invention relate to a constructing a gene interaction network. Tools are provided to compute a gene relationship measure based upon cellular images, and to rank image collections having a similar morphology. The ranking is based upon capturing similarity within the ranked collection by modeling a three dimensional shape of a cellular image stack. The graph is constructed for related images stacks. Nodes in the graph represent genes, and edges drawn between the nodes represent corresponding image stacks in a commonly ranked list. Accordingly, the graphical representation mathematically and visually connects respective genes.

主权项

1. A system comprising:
a processor in communication with memory and data storage; the processor programmed to construct a gene interaction network including the processor to: generate a cellular image stack via RNAi screening of a set of genes, wherein the generation comprises receipt of recorded images of groups of cells associated with a first set of genes, wherein each of the groups is received by knocking down one or more genes from a second set of genes and wherein the first set of genes and the second set of genes include common genes; store the generated cellular image stack in the memory; create a set of three-dimensional feature vectors from the stored cellular image data; compare the stored cellular image data, based on the set of feature vectors, and produce a gene relationship measure including computing the gene relationship measure using a morphological similarity measure between the corresponding groups of cells reflecting structural similarities between the images being compared; convert the gene relationship measure into a ranking of images with similar morphology, including modeling a three-dimensional shape of the cellular image stack produced by knocking down respective genes; construct a graph for related image stacks, draw edges between all nodes of the graph based on the image ranking being in a commonly ranked list, and apply a weight to each edge based on node similarity, wherein each gene is a node in the graph; and output the graph to the processor.