Data Exploration, Quality Control and Testing in Single-Cell qPCR-Based\n Gene Expression Experiments

Abstract

Cell populations are never truly homogeneous; individual cells exist in\nbiochemical states that define functional differences between them. New\ntechnology based on microfluidic arrays combined with multiplexed quantitative\npolymerase chain reactions (qPCR) now enables high-throughput single-cell gene\nexpression measurement, allowing assessment of cellular heterogeneity. However\nvery little analytic tools have been developed specifically for the statistical\nand analytical challenges of single-cell qPCR data. We present a statistical\nframework for the exploration, quality control, and analysis of single-cell\ngene expression data from microfluidic arrays. We assess accuracy and\nwithin-sample heterogeneity of single-cell expression and develop quality\ncontrol criteria to filter unreliable cell measurements. We propose a\nstatistical model accounting for the fact that genes at the single-cell level\ncan be on (and for which a continuous expression measure is recorded) or\ndichotomously off (and the recorded expression is zero). Based on this model,\nwe derive a combined likelihood-ratio test for differential expression that\nincorporates both the discrete and continuous components. Using an experiment\nthat examines treatment-specific changes in expression, we show that this\ncombined test is more powerful than either the continuous or dichotomous\ncomponent in isolation, or a t-test on the zero-inflated data. While developed\nfor measurements from a specific platform (Fluidigm), these tools are\ngeneralizable to other multi-parametric measures over large numbers of events.\n