SCAFE: a software suite for analysis of transcribed cis-regulatory elements in single cells

Abstract

Motivation Cell type-specific activities of cis-regulatory elements (CRE) are central to understanding gene regulation and disease predisposition. Single-cell RNA 5’end sequencing (sc-end5-seq) captures the transcription start sites (TSS) which can be used as a proxy to measure the activity of transcribed CREs (tCREs). However, a substantial fraction of TSS identified from sc-end5-seq data may not be genuine due to various artifacts, hindering the use of sc-end5-seq for de novo discovery of tCREs. Results We developed SCAFE—Single Cell Analysis of Five-prime Ends—a software suite that processes sc-end5-seq data to de novo identify TSS clusters based on multiple logistic regression. It annotates tCREs based on the identified TSS clusters and generates a tCRE-by-cell count matrix for downstream analyses. The software suite consists of a set of flexible tools that could either be run independently or as pre-configured workflows. Availability SCAFE is implemented in Perl and R. The source code and documentation are freely available for download under the MIT License from https://github.com/chung-lab/SCAFE. Docker images are available from https://hub.docker.com/r/cchon/scafe. The submitted software version and test data are archived at https://doi.org/10.5281/zenodo.7023163 and https://doi.org/10.5281/zenodo.7024060 respectively. Supplementary information Supplementary data are available at Bioinformatics online.

Atlas of nascent RNA transcripts reveals enhancer to gene linkages

“…However, within the context of a single tissue we observe fewer bidirectional regions linked to a gene (median = 2-8; Supplemental Figure 19). This estimate is on par with other estimates of number of enhancers linked to a gene [47][48][49][50][51][52][53][54]. In the other direction, the median number of genes assigned to a bidirectional transcript across all tissues was four, implying that a single bidirectional has only a few potential gene targets (Supplementary Figures 18C and D).…”

A single-cell atlas of transcribedcis-regulatory elements in the human genome

“…Utilizing our single-cell data, we built a tCRE atlas comprising 81,829 proximal (P-tCREs) and 96,400 distal (D-tCREs) elements (Methods; Moody et al, 2022; Supplementary Table 2). The majority of these tCREs—94.3% of P-tCREs and 88.2% of D-tCREs were supported by candidate CREs from external epigenomic datasets from ENCODE (ENCODE Project Consortium et al ., 2020) and a sc-ATAC atlas (Zhang et al ., 2021), affirming the validity of our tCREs ( Fig.…”

“…To identify tCRE in the atlas data, the SCAFE v1.0.0 (Moody et al ., 2022) pipeline was applied to define and annotate tCREs. Briefly, for each library, the single nucleotide resolution 5’Cap TSS (CTSS) signals, including the number of UMI with unencoded Gs, were extracted from the alignment bam files generated from cellranger .…”

“…The CTSS signals for all libraries of each “project” (as listed in Supplementary Table 1) were aggregated and used to define a set of TSS clusters for each project. For each project, the pooled CTSS signals were clustered using paraclu within SCAFE (Moody et al, 2022) using default parameters, with a cutoff set to ≥3 UMI of encoded-G supported TSS per TSS cluster. TSS clusters that are potentially strand invasion artifacts were removed (Moody et al, 2022).…”

“…For each project, the pooled CTSS signals were clustered using paraclu within SCAFE (Moody et al, 2022) using default parameters, with a cutoff set to ≥3 UMI of encoded-G supported TSS per TSS cluster. TSS clusters that are potentially strand invasion artifacts were removed (Moody et al, 2022). The remaining TSS clusters were further filtered using a logistic regression classifier trained with matched sc-5’-RNASeq and ATAC-Seq data implemented in SCAFE (Moody et al, 2022) at the logistic probability cutoff of 0.9.…”