EVALUATION OF CURRENT MOST SUCCESSFUL METHODS FOR DECONVOLUTION OF CELL MIXTURES BASED ON  WEIGHTED LEAST SQUARES REGRESSION APPROACHES

Natalia Alonso-Moreda; José Manuel Sánchez-Santos; Javier De Las Rivas

Paper Title

EVALUATION OF CURRENT MOST SUCCESSFUL METHODS FOR DECONVOLUTION OF CELL MIXTURES BASED ON WEIGHTED LEAST SQUARES REGRESSION APPROACHES

Authors

Natalia Alonso-Moreda
José Manuel Sánchez-Santos
Javier De Las Rivas

Modality

RIABIO

Subject area

RNA and transcriptomics

Publishing Date

21/11/2024

Country of Publishing

Brazil | Brasil

Language of Publishing

Inglês

Paper Page

https://www.even3.com.br/anais/xmeeting-2024/832623-evaluation-of-current-most-successful-methods-for-deconvolution-of-cell-mixtures-based-on--weighted-least-squares

ISBN

978-65-272-0843-3

Keywords

Bioinformatics, deconvolution, transcriptomics, RNA-Seq, scRNA-seq, oncology data, tumor microenvironment

Summary

Tumors participate in a complex Tumor Microenvironment (TME) that involves a dynamic network of cellular interactions between malignant, immune, and stromal cells. The presence and abundance of specific cell types have been shown to influence tumor progression and treatments response. Traditionally, some experimental techniques have been used to infer cellular composition. However, they have some limitations related to the time and costs involved, as well as a quite reduced number of cell types tested. Hence, computational techniques known as deconvolution methods have been developed to decompose signal mixtures (bulk) and infer the cellular composition contained in biological samples that use reference gene profiles based on Single-Cell RNA-Seq (scRNA-Seq) data or predefined signature matrices. They then apply a specific mathematical approximation to estimate the cell proportions, using both the bulk and reference profiles. Currently, most of the highly successful approaches are based on the use of "least squares regression". Some commonly used are Ordinary Least Squares (OLS), Super Vector Regression (SVR), or Non-Negative Least Squares (NNLS). In addition, some others also add a gene weighting factor to modulate the regression calculation. In this work, we focused on methods that rely on scRNA-Seq reference profile signatures and evaluated the currently most successful methods for deconvolution of cell mixtures based on "weighted least squares regression" approaches. The selected deconvolution algorithms were: Dampened Weighted Least Squares (DWLS); MuSiC (Multi-subject Single-Cell deconvolution), which is based on Weighted Non Negative Least Squares (W-NNLS); and Single-cell RNA Quantity-Informed Deconvolution (SQUID), which is a simplification of DWLS. As can be seen, they are all based on similar mathematical approximations, but with slight modifications. It is important to understand and evaluate the results derived from these variations to assess the methods well and to find an optimized methodology for deconvolution of complex cell mixtures, especially in human samples.

Title of the Event: 20º Congresso Brasileiro de Bioinformática: X-Meeting 2024
City of the Event: Salvador
Title of the Proceedings of the event: X-Meeting presentations
Name of the Publisher: Even3
Means of Dissemination: Meio Digital

How to cite

ALONSO-MOREDA, Natalia; SÁNCHEZ-SANTOS, José Manuel; RIVAS, Javier De Las. EVALUATION OF CURRENT MOST SUCCESSFUL METHODS FOR DECONVOLUTION OF CELL MIXTURES BASED ON WEIGHTED LEAST SQUARES REGRESSION APPROACHES.. In: X-Meeting presentations. Anais...Salvador(BA) Hotel Deville Prime, 2024. Available in: https//www.even3.com.br/anais/xmeeting-2024/832623-EVALUATION-OF-CURRENT-MOST-SUCCESSFUL-METHODS-FOR-DECONVOLUTION-OF-CELL-MIXTURES-BASED-ON--WEIGHTED-LEAST-SQUARES. Access in: 18/05/2025