Collective properties of cellular identity: a computational approach

Bradly Alicea

DOI: https://doi.org/10.48550/arXiv.1302.0826

2013-02-04

Quantitative Methods

Abstract:Cell type (e.g. pluripotent cell, fibroblast) is the end result of many complex processes that unfold due to evolutionary, developmental, and transformational stimuli. A cell's phenotype and the discrete, a priori states that define various cell subtypes (e.g. skin fibroblast, embryonic stem cell) are ultimately part of a continuum that may predict changes and systematic variation in cell subtypes. These features can be both observable in existing cellular states and hypothetical (e.g. unobserved). In this paper, a series of approaches will be used to approximate the continuous diversity of gene expression across a series of pluripotent, totipotent, and fibroblast cellular subtypes. We will use a series of previously-collected datasets and analyze them using three complementary approaches: the computation of distances based on the subsampling of diversity, assessing the separability of individual genes for a specific cell line both within and between cell types, and a hierarchical soft classification technique that will assign a membership value for specific genes in specific cell types given a number of different criteria. These approaches will allow us to assess the observed gene-expression diversity in these datasets, as well as assess how well a priori cell types characterize their constituent populations. In conclusion, the application of these findings to a broader biological context will be discussed.