(199a) Unbiased Algorithms for Analysis of Multidimensional Phenotypic Profiles Reveal Hidden Alleles and Link Phenotypes to Genotypes

Authors: 
San-Miguel, A., Georgia Institute of Technology
Lu, H., Georgia Institute of Technology

New approaches to biological studies from a systems-level perspective, aim to find relationships between the ensemble of molecular players required for biological functions. These new approaches necessitate that experimental platfrms focus on large data sets and multi-parametric characterization. At the same time, imaging of fluorescent markers to study spatiotemporal expression of proteins and biomolecules has become ubiquitous in biological experiments. Large imaging data sets are, however, difficult to analyze in an unbiased manner. Moreover, the stochastic nature of gene expression and environmental noise, make the identification of true subtle phenotypic differences exceedingly difficult. One approach to tackle these large data sets is to rely on mathematical and statistical approaches to extract relevant information, particularly in problems where phenotypes are subtle and the characteristic features are not known a priori.

Here, we take advantage of unbiased algorithms to extract multidimensional data from content-rich images of synaptic patterns in C. elegans. We focus on identifying relevant features, and “meta-features” that provide information beyond the qualitatively identifiable characteristics. We applied unbiased algorithms to identify whether true differences existed between populations of animals with different genetic background, and where these differences come from. We applied these tools to animals isolated in a genetic screen, with the aim of finding genetic perturbations that caused synaptic patterning phenotypes invisible to the human eye. Without prior knowledge of the phenotypic differences, we are able to establish whether these are true mutants and where these differences may reside.. Phenotypic clustering tools suggested putative altered biological functions of the isolated “invisible” alleles. These tools are applicable to any multidimensional data set, and enables unbiased analysis of the truly relevant features that characterize biological populations.