Skip to Main Content
Phenotyping cells and tracking their functional states are key tasks in cell biology and molecular medicine. Current cell classification methods are idiosyncratic to specific fields and based on ad hoc discovery of presumed univariate markers. We propose a general theory of phenotyping based on broadly distributed multivariate markers as the metrics of classification and standard pattern recognition algorithms as the method of class discovery. We present a real-world test case based on the vertebrate retina and demonstrate that pattern recognition methods can extract singular populations of neurons from complex heterocellular arrays: populations visualized solely as elements in a micromolecular N-space. The applications of this computational approach to cell phenotyping range from phylogenetics to drug discovery to environmental monitoring.