GAGAM v1.2: An Improvement on Peak Labeling and Genomic Annotated Gene Activity Matrix Construction

Abstract

Single-cell Assay for Transposase-Accessible Chromatin using sequencing (scATAC-seq) is rapidly becoming a powerful technology for assessing the epigenetic landscape of thousands of cells. However, the sparsity of the resulting data poses significant challenges to their interpretability and informativeness. Different computational methods are available, proposing ways to generate significant features from accessibility data and process them to obtain meaningful results. Foremost among them is the peak calling, which interprets the raw scATAC-seq data generating the peaks as features. However, scATAC-seq data are not trivially comparable with single-cell RNA sequencing (scRNA-seq) data, an increasingly pressing challenge since the necessity of multimodal experiments integration. For this reason, this study wants to improve the concept of the Gene Activity Matrix (GAM), which links the accessibility data to the genes, by proposing an improved version of the Genomic-Annotated Gene Activity Matrix (GAGAM) concept. Specifically, this paper presents GAGAM v1.2, a new and better version of GAGAM v1.0. GAGAM aims to label the peaks and link them to the genes through functional annotation of the whole genome. Using genes as features in scATAC-seq datasets makes different datasets comparable and allows linking gene accessibility and expression. This link is crucial for gene regulation understanding and fundamental for the increasing impact of multi-omics data. Results confirm that our method performs better than the previous GAMs and shows a preliminary comparison with scRNA-seq data.

Meta-analysis of gene activity (MAGA) contributions and correlation with gene expression, through GAGAM

“…Beyond these simplistic models, other approaches aim to comprise more accessible genomic regions and their impact on overall GA. Specifically, this work employs GAGAM since it uses curated genomic annotations to functionally label the peaks and, consequentially, associates them with the genes through a simple model [15].…”

“…Finally, GAGAM sums all the contributions weighted by model-specific weights to obtain the final GA matrix. In the GAGAM implementation introduced in [15], simple binary weights are used, but a better understanding of the contribution of the three matrices could help fine-tune them. The final model of GAGAM activity translates the simple accessibility data from scATAC-seq into a score representing the overall accessibility of the gene and its potential to be expressed.…”

“…The scATAC-seq part of the dataset has a total of 115,179 peaks as features, while the scRNA-seq part has 36,601 genes. The tools employed to process and elaborate the data are GAGAM (the focus of this paper, accessible from [15]) and Seurat [4]. The latter is one of the most wellknown and highly-utilized single-cell pipelines.…”

“…Our previous work [15] qualitatively addressed this type of investigation to broadly study Activity-Expression (A-E) patterns for specific genes. This paper applies a quantitative approach.…”

“…A promising approach to solve this problem is the Genomic-Annotated Gene Activity Matrix (GAGAM) approach [15] [16]. In GAGAM, the association between genomic regions and accessible genes relies on a genomic model based on genomic annotations.…”