Calculating the quality of public high-throughput sequencing data to obtain a suitable subset for reanalysis from the Sequence Read Archive

Abstract

It is important for public data repositories to promote the reuse of archived data. In the growing field of omics science, however, the increasing number of submissions of high-throughput sequencing (HTSeq) data to public repositories prevents users from choosing a suitable data set from among the large number of search results. Repository users need to be able to set a threshold to reduce the number of results to obtain a suitable subset of high-quality data for reanalysis. We calculated the quality of sequencing data archived in a public data repository, the Sequence Read Archive (SRA), by using the quality control software FastQC. We obtained quality values for 1 171 313 experiments, which can be used to evaluate the suitability of data for reuse. We also visualized the data distribution in SRA by integrating the quality information and metadata of experiments and samples. We provide quality information of all of the archived sequencing data, which enable users to obtain sufficient quality sequencing data for reanalyses. The calculated quality data are available to the public in various formats. Our data also provide an example of enhancing the reuse of public data by adding metadata to published research data by a third party.

Keywords: high-throughput sequencing; sequencing quality; public data; database.

Figure 1:

Performed sequencing experiments and sequenced samples of public data for quality calculation. (a) Bar plot of the top 20 library strategies. Values are categorical, retrieved from metadata described by the data submitter. (b) Bar plot of the top 20 sequenced sample organisms. Taxonomy information is retrieved from the NCBI taxonomy database and declared by the data submitter. (c) Bar plot of sequencing instrument models.

Figure 2:

Data distribution in a public data repository by sequencing quality. (a, b) Histogram of sequencing throughput (a) and one color-coded by library source (b). (c, d) Histogram of base call accuracy (c) and one color-coded by instrument manufacturer (d).

Figure 3:

Human data distribution for each library strategy. (a–d) Histograms separated by the top six library strategies. Data distribution is by the total number of sequences (a), median read length (b), sequencing throughput (c), and median base call accuracy (d) per experiment.

Figure 4:

Change of data distribution by sequencing quality over time. (a, b) Box plots separated by the top six library strategies, showing quarterly change. Data distribution is by the sequencing throughput (a) and median base call accuracy (b) per experiment. The numbers in the plots indicate the numbers of samples in a row. The lines connecting the boxes indicate changes of mean value.