Li, Shaoyu

Abstract This study benchmarks the robustness and resilience of computational deconvolution methods for estimating cell-type proportions in bulk tissues, with a focus on comparing reference-based and reference-free methods. Robustness is evaluated by generating in silico pseudo-bulk tissue RNA sequencing data from cell-level gene expression profiles derived from four different tissue types, with simulated cellular composition at varying levels of heterogeneity. To assess resilience, we intentionally alter single-cell RNA profiles to create pseudo-bulk tissue RNA-seq data. Deconvolution estimates are compared with ground truth using Pearson’s correlation coefficient, root mean squared deviation, and mean absolute deviation. The results show that reference-based methods are more robust when reliable reference data are available, whereas reference-free methods excel in scenarios lacking suitable reference data. Furthermore, variations in cell-level transcriptomic profiles and cell composition have emerged as critical factors influencing the performance of deconvolution methods. This study provides significant insights into the factors affecting bulk tissue deconvolution performance, which are essential for guiding users and advancing the development of more powerful and reliable algorithms in the future.

<p style='text-indent:20px;'>Complete deconvolution analysis for bulk RNA-seq data is important and helpful to distinguish whether the differences of disease-associated GEPs (gene expression profiles) in tissues of patients and normal controls are due to changes in cellular composition of tissue samples, or due to GEPs changes in specific cells. One of the major techniques to perform complete deconvolution is nonnegative matrix factorization (NMF), which also has a wide-range of applications in the machine learning community. However, the NMF is a well-known strongly ill-posed problem, so a direct application of NMF to RNA-seq data will suffer severe difficulties in the interpretability of solutions. In this paper, we develop an NMF-based mathematical model and corresponding computational algorithms to improve the solution identifiability of deconvoluting bulk RNA-seq data. In our approach, we combine the biological concept of marker genes with the solvability conditions of the NMF theories, and develop a geometric structures guided optimization model. In this strategy, the geometric structure of bulk tissue data is first explored by the spectral clustering technique. Then, the identified information of marker genes is integrated as solvability constraints, while the overall correlation graph is used as manifold regularization. Both synthetic and biological data are used to validate the proposed model and algorithms, from which solution interpretability and accuracy are significantly improved.</p>