ARIC: accurate and robust inference of cell type proportions from bulk gene expression or DNA methylation data

Abstract

Quantifying cell proportions, especially for rare cell types in some scenarios, is of great value in tracking signals associated with certain phenotypes or diseases. Although some methods have been proposed to infer cell proportions from multicomponent bulk data, they are substantially less effective for estimating the proportions of rare cell types which are highly sensitive to feature outliers and collinearity. Here we proposed a new deconvolution algorithm named ARIC to estimate cell type proportions from gene expression or DNA methylation data. ARIC employs a novel two-step marker selection strategy, including collinear feature elimination based on the component-wise condition number and adaptive removal of outlier markers. This strategy can systematically obtain effective markers for weighted $\upsilon$-support vector regression to ensure a robust and precise rare proportion prediction. We showed that ARIC can accurately estimate fractions in both DNA methylation and gene expression data from different experiments. We further applied ARIC to the survival prediction of ovarian cancer and the condition monitoring of chronic kidney disease, and the results demonstrate the high accuracy and robustness as well as clinical potentials of ARIC. Taken together, ARIC is a promising tool to solve the deconvolution problem of bulk data where rare components are of vital importance.

Deconer: A comprehensive and systematic evaluation toolkit for reference-based cell type deconvolution algorithms using gene expression data

“…Accurately estimating the proportions of specific cell types, especially rare cell types in certain circumstances, is of crucial importance in uncovering their significance in specific phenotypes or diseases [4,20]. For instance, precise estimation of the proportion for tumor-infiltrating lymphocytes (TILs) plays a vital role in predicting clinical prognosis and developing personalized treatment strategies [58–60].…”

“…For scaden and TAPE, there are certain differences in the way they generate training data. Specifically, scaden directly samples and normalizes from a uniform distribution, which is similar to the approach used in some other studies [4,15]. On the other hand, the TAPE method generates random proportions using the Dirichlet distribution, which can better simulate real-world scenarios when there are certain prior proportions [45].…”

“…The training data was solely used to generate the external reference, while the test data was utilized to produce samples for assessment. Evaluation samples were directly generated using the same linear mixing approach as ARIC [4]. In brief, numerical linear combinations were performed after generating cell proportions according to the requirements of different evaluation criteria.…”

“…The rapid development of high-throughput sequencing technologies, such as RNA-seq and scRNA-seq, provide us unprecedented opportunities to unravel the global gene expression in cells [1,2]. In multicellular organisms, nearly all physiological and pathological processes involve multiple types of cells, each playing a unique role in specific mechanisms [3,4]. Quantifying cell proportions is essential for tracking signals associated with certain phenotypes or diseases.…”

stMMR: accurate and robust spatial domain identification from spatially resolved transcriptomics with multi-modal feature representation

“…The researchers only need to duplicate the gene expression module and then apply similar methods for embedding features within and between modalities. Moreover, the joint representation obtained through stMMR can also be applied to other tasks, such as cell type deconvolution [82][83][84][85]. This application requires a process similar to methods like scaden [86], where a neural network is connected to the joint representation.…”