Compensation has long been one of the most perplexing aspects of cytometry, with the most critical requirement being pristine compensation controls for each and every parameter in an experiment. With our industry seeing consistent exponential growth in the number of parameters collected per experiment, there has been a corresponding exponential rise in the effort required to create appropriate compensation. AutoSpill makes cytometry easier by relaxing the need for perfect controls and eliminating many of the steps needed for processing those controls.
Rather than requiring exemplar ‘positive’ and ‘negative’ gates for each florescent probe, AutoSpill uses a robust linear regression approach to approximate the spillover in an experiment. In traditional compensation we assume that the median fluorescent intensity (MFI) of a population assumed to be negative for a particular florescent probe will be equal to the MFI of cells that have bound the probe that the sample is a positive control for, for all other colors. We then create a spillover matrix by solving a system of linear equations so that this is true for all parameters. AutoSpill assumes instead that the slope of a regression line fit to the data will be zero for properly compensated data and creates a spillover matrix that solves a system of linear equations that makes this true.
Further, AutoSpill uses an optimization routine to alleviate the need to manually adjust the compensation matrix in most cases, by iteratively applying this process to the resulting compensated data to refine the matrix.
AutoSpread is a measure of the spreading error associated with the matrix that that is compatible with AutoSpill. The traditional measure of the goodness of a matrix is the Spillover Spreading matrix (SSM), which uses statistics calculated on the ‘positive’ and ‘negative’ gates. As AutoSpill uses no gates an alternate approach was needed. The spread in fluorescence for any (compensated or uncompensated) channel/dye grows linearly with the fluorescence level, and therefore the coefficients of the SSM can be estimated with linear regression. The AutoSpread approach bins the single stain control data as the input to calculate spread without gates.
Uses and Limitations
Overall, the combination of these two tools makes compensation both easier and more robust. For more information on this new approach, see the Nature Webinar on AutoSpill / AutoSpread.