GCA

"DeCompress" is a software tool to address the complexities and challenges of analyzing mRNA expression data derived from bulk tissue samples. These samples are typically characterized by a mixture of different cell types, which can obscure the biological signals of interest due to cell-type heterogeneity. This problem is particularly pronounced when using targeted mRNA expression panels, favored in academic and clinical settings for their cost-effectiveness and high sensitivity, especially with archived samples. However, these panels focus on a limited number of genes (up to 800), further complicating the analysis due to the restricted feature space.

To overcome these limitations, DeCompress introduces a semi-reference-free deconvolution approach that ingeniously expands the feature space available for analysis. How DeCompress works:

• Semi-Reference-Free Deconvolution: Unlike purely reference-free methods, which do not use any cell-type-specific expression references (often due to their unavailability), DeCompress adopts a semi-reference-free strategy. This approach allows for including some reference data to inform the deconvolution process, thereby improving accuracy.
• Leveraging External Reference Data: DeCompress utilizes RNA-seq or microarray data from tissue similar to that being studied to expand the feature space of targeted panels artificially. DeCompress achieves this through a process known as compressed sensing, which essentially allows for reconstructing a signal (in this case, gene expression profiles) from a few observations.
• Ensemble Reference-Free Deconvolution: Once the feature space is expanded, DeCompress performs ensemble deconvolution on this enhanced dataset. This process estimates the proportions of different cell types within the sample and their gene signatures, which are crucial for understanding the biological context of the data.

Features of GCA:

• Comprehensive Analysis for Pedigree and Genomic Data: GCA is specifically designed to perform connectedness analysis on both pedigree and genomic data.
• Collection of Connectedness Statistics: The software implements a vast array of connectedness statistics. These statistics can be based on prediction error variance or the variance of unit effect estimates, providing users with a detailed understanding of the genetic connectedness in their data.