SAMStat: monitoring biases in next generation sequencing data

Abstract

Motivation: The sequence alignment/map format (SAM) is a commonly used format to store the alignments between millions of short reads and a reference genome. Often certain positions within the reads are inherently more likely to contain errors due to the protocols used to prepare the samples. Such biases can have adverse effects on both mapping rate and accuracy. To understand the relationship between potential protocol biases and poor mapping we wrote SAMstat, a simple C program plotting nucleotide overrepresentation and other statistics in mapped and unmapped reads in a concise html page. Collecting such statistics also makes it easy to highlight problems in the data processing and enables non-experts to track data quality over time.

Results: We demonstrate that studying sequence features in mapped data can be used to identify biases particular to one sequencing protocol. Once identified, such biases can be considered in the downstream analysis or even be removed by read trimming or filtering techniques.

Availability: SAMStat is open source and freely available as a C program running on all Unix-compatible platforms. The source code is available from http://samstat.sourceforge.net.

Contact: timolassmann@gmail.com.

Fig. 1.

A selection of SAMStat's html output. (a) Mapping statistics. More than half of the reads are mapped with a high mapping accuracy (red) while 9.9% of the reads remain unmapped (black). (b) Barcharts showing the distribution of mismatches and insertions along the read for alignments with the highest mapping accuracy [shown in red in (a)]. The colors indicate the mismatched nucleotides found in the read or the nucleotides inserted into the read. (c,d and e) Frequency of mismatches at the start of reads with mapping accuracies 1e⁻³ ≤ P < 1e⁻², 1e⁻² ≤ P < 0.5 and 0.5 ≤ P < 1, respectively (shown in orange, yellow and blue in panel a). The fraction of mismatches involving G's at position 2–5 increases. (f) Percentage of ‘GG’ dinucleotides at positions 1–5 in reads split up by mapping quality intervals. The background color highlights large percentages. The first and last row for nucleotides ‘GT’ and ‘GC’ are shown for comparison.