COT: an efficient and accurate method for detecting marker genes among many subtypes

Abstract

Motivation Ideally, a molecularly distinct subtype would be composed of molecular features that are expressed uniquely in the subtype of interest but in no others—so-called marker genes (MGs). MG plays a critical role in the characterization, classification or deconvolution of tissue or cell subtypes. We and others have recognized that the test statistics used by most methods do not exactly satisfy the MG definition and often identify inaccurate MG. Results We report an efficient and accurate data-driven method, formulated as a Cosine-based One-sample Test (COT) in scatter space, to detect MG among many subtypes using subtype expression profiles. Fundamentally different from existing approaches, the test statistic in COT precisely matches the mathematical definition of an ideal MG. We demonstrate the performance and utility of COT on both simulated and real gene expression and proteomics data. The open source Python/R tool will allow biologists to efficiently detect MG and perform a more comprehensive and unbiased molecular characterization of tissue or cell subtypes in many biomedical contexts. Nevertheless, COT complements not replaces existing methods. Availability and implementation The Python COT software with a detailed user’s manual and a vignette are freely available at https://github.com/MintaYLu/COT. Supplementary information Supplementary data are available at Bioinformatics Advances online.

ABDS: a bioinformatics tool suite for analyzing biologically diverse samples

“…We evaluated the performance of MGpI and eCOT in comparison with representative or standard peer methods 5,6 . The evaluation does not include uniHM because it is a subjective visualization tool.…”

“…Ideally, a signature gene among molecularly distinct groups would be either uniquely expressed or silent in the group of interest but in no others 4 . However, test statistics used by most existing methods do not satisfy exactly this signature definition and are theoretically prone to detecting imprecise signatures 5 . Furthermore, while a typical heatmap design is visually effective, the common reference origin for expression measurements is altered by the classical standardization, with zero-expression replaced by floating negative values for different genes.…”

“…Here we present ABDS tool suite assembled specifically for analyzing biologically diverse samples. Open-source R package includes three fundamental and interrelated analytic tools, namely, mechanism-integrated group-wise pre-imputation (MGpI), extended cosinebased one-sample test (eCOT) 5 , and unified heatmap design (uniHM). Collectively, we propose a hybrid imputation strategy to impute informative missingness associated with signature genes (SG), a cosine-score test to detect downregulated signature genes (DSG), and a unified heatmap design to comparably display multiple differential groups (Fig.…”

Machine Learning for Synthetic Data Generation: a Review

“…In order to protect health information and improve reproducibility in research, synthetic data has drawn mainstream attention in the healthcare industry [54,55]. Many labs and companies have harnessed the tools of big data and advanced computation tools to produce large quantities of synthetic data [56]. Modeled after patient data, synthetic data generation is essential to understanding diseases while maintaining patient confidentiality and privacy simultaneously [57].…”

Cosbin: cosine score-based iterative normalization of biologically diverse samples

“…As a novel normalization-by-testing strategy ( Evans et al , 2018 ), the Cosbin framework makes no assumption that the total expression is the same or that differential expression across differential experimental conditions is approximately symmetrical and thus complements rather than replaces existing methods ( Evans et al , 2018 ; Johnson and Krishnan 2022 ). One additional benefit of the Cosbin tool is the concurrent detection of marker genes (MGs) and iCEGs ( Lu et al , 2022 ) ( Supplementary Information ).…”