About DASHR v2.0

DASHR implementation

The DASHR database was implemented using MySql database engine, while all the scripts were written in PHP (website), AWK and BASH scripting languages. R (v3.3.3) is used for plotting and visualization. DASHR v2.0 is implemented on Amazon Cloud (m3.large instance with Intel Xeon E5-2670 2.5GHz CPUs and 8GB RAM) storing 10TB of raw sequencing data and with processed data stored in MySQL. The database schema can be viewed here.

All small RNA-seq datasets were processed using SPAR pipeline (SPAR manuscript https://doi.org/10.1093/nar/gky330, SPAR bitbucket repository SPAR repository).

Methods - Small-RNA-sequencing and annotation pipeline

Annotation resources

DASHR integrates multiple existing annotation resources for small non-coding RNAs including miRNA annotations from miRBase (v19); snRNA, snoRNA, scRNA and rRNA annotations are from UCSC genome browser; and GENCODE; tRNA annotations tRNAdb; and piRNA annotations from NCBI. The current annotation in DASHR also includes tRNA fragment annotations that are created based on the 50nt sequences upstream (5p) and downstream (3p) of the known tRNA genes. Information from NCBI, HGNC, RefSeq and UCSC was used to build the ID conversion table for each processed and mature sncRNA entry in DASHR.

Data collection and curation of smRNA-seq datasets

We manually curated Illumina smRNA-seq datasets from Gene Expression Omnibus (GEO) and Sequence Read Archive (SRA). These datasets were obtained from non-diseased human tissues and cell types and were generated for studying or profiling small non-coding RNAs. The full list of included samples (smRNA-seq libraries) categorized by tissue and cell type is shown in the Data page.

Categorizing datasets into tissues and cell types

We then manually categorized the smRNA-seq samples into different groups of tissues and cell types, stratified by their experimental sources (i.e. Study ID found in Data summary page).

Processing smRNA-seq datasets

After curation and categorization, we then used SPAR, a modified pipeline built upon CoRAL to process the smRNA-seq datasets and generate sncRNA expression levels. The pipeline can be summarized in three parts: 1) processing and quality control; 2) segmentation and quantification; and 3) annotation.

Processing and Quality control (trimming, alignment)

We first identified the correct adapter (Illumina adapter column found in Data summary page) and trimmed the sequencing reads. We then mapped the trimmed reads to a standardized version of the human reference genome GRCh37/hg19 and GRCh38/hg38. The reads were aligned using STAR allowing for multi-mapping. Over 80% of the trimmed reads were mapped to the human genome on average per dataset (Data summary page details trimming and alignment statistics for each smRNA-seq dataset).

Segmentation and quantification of sncRNA loci

As one of our goals was to identify mature sncRNA products (i.e. peaks with evidence of specific processing patterns, e.g. with low 5p read entropy) and quantify their expression levels in addition to precursor sncRNA gene expression levels, we developed a customized approach to identify peaks with evidence of specific processing for mature sncRNA products in base pair resolution. After identifying the mature sncRNA locations, we then quantified the number of reads falling within these regions as read counts for each sncRNA. To enable comparison across tissues, we took into account the library size information (i.e. number of mapped reads) for each of the sequencing experiments (found in Data summary page) and reported the read count in "reads per million" (RPM).

Annotation of mature RNA products

We overlapped each peak with evidence of specific processing (identified in the previous step) with DASHR annotations (see Annotation resources). Each peak was assigned to its precursor RNA class and sncRNA gene.

Outputs

Summary table

The summary table contains annotation information, including the sequence, genomic coordinates, and structural information for the selected locus. This table also includes summary statistics for expression information, highlighting the total number of experiments with observed expression, total raw read count and total reads count in reads per million (RPM).

Expression profiles

For any set of the specific tissues/cell types, users can view expression profiles or read coverage profiles representing detailed expression information for any sncRNA locus across the set of tissues/cell types. There are also links to the UCSC genome browser, allowing users to view the genomic locus with mapped sequencing data across all tissues.

The expression profile plot shows the reads in RPM across all tissues in DASHR for a given locus. Users can choose to display the expression profile in a bar-plot or heatmap format, in which the RPM values are shown.

For anyone (or all) chosen specific tissues/cell types, users can view the detailed expression information represented by expression profiles or read coverage profiles.

The read coverage plot available for a single tissue or the entire set of tissues, displays the number of raw reads from the chosen tissue along the genomic coordinates of the locus. The sequence is displayed underneath the plot.