Brains are comprised of many types of cells. Major cell types in human brains include endothelial, neuronal, astrocytes, glia, oligodendrocytes, and many more. Each cell type is involved in several biological processes like signaling, blood flow, immune interactions, etc. Such processes typically involve complex interaction among several distinct types. Important to understanding such processes is understanding cell type heterogeneity. Cell-type composition of a brain can be indicative of diseases or dysfunctions like Alzheimers or ALS. Methods to estimate cell type proportions from gene expression data, known as cell-type deconvolution, have been extensively studied. These methods estimate cell-type proportion in a mixture by comparing its gene expressions to references of various cell types. Unfortunately, these references often contain misleading technical and biological effects that impair accuracy. For example, there may be batch effects across technologies or physical processing labs. We explore methods for attenuating unwanted technical and biological factors in the context of deconvolution of human brain samples.