swCAM: estimation of subtype-specific expressions in individual samples with unsupervised sample-wise deconvolution

Abstract

MotivationComplex biological tissues are often a heterogeneous mixture of several molecularly distinct cell or tissue subtypes. Both subtype compositions and expressions in individual samples can vary across different biological states or conditions. Computational deconvolution aims to dissect patterns of bulk gene expression data into subtype compositions and subtype-specific expressions. Typically, existing deconvolution methods can only estimate averaged subtype-specific expressions in a population, while detecting differential expressions or co-expression networks in particular subtypes requires unique subtype expression estimates in individual samples. Different from population-level deconvolution, however, individual-level deconvolution is mathematically an underdetermined problem because there are more variables than observations. ResultsWe report a sample-wise Convex Analysis of Mixtures (swCAM) method that can estimate subtype proportions and subtype-specific expressions in individual samples from bulk tissue transcriptomes. We extend our previous CAM framework to include a new term accounting for between-sample variations and formulate swCAM as a nuclear-norm and{ell} 2,1-norm regularized matrix factorization problem. We determine hyperparameter values using a cross-validation scheme with random entry exclusion and obtain a swCAM solution using an efficient alternating direction method of multipliers. The swCAM is implemented in open-source R scripts. Experimental results on realistic simulation data show that swCAM can accurately estimate subtype-specific expressions in individual samples and successfully extract co-expression networks in particular subtypes that are otherwise unobtainable using bulk expression data. Application of swCAM to bulk-tissue data of 320 samples from bipolar disorder patients and controls identified changes in cell proportions, expression and coexpression modules in patient neurons. Mitochondria related genes showed significant changes suggesting an important role of energy dysregulation in bipolar disorder. Availability and implementationThe R Scripts of swCAM is freely available at https://github.com/Lululuella/swCAM. A users guide and a vignette are provided. Contactyuewang@vt.edu Supplementary informationSupplementary data are available at Bioinformatics online.