Variant calling in human whole genome/exome sequencing data

Edit me

Step 1.3: Removal of very short reads

Once the adapter remnants and low quality ends have been trimmed, some reads may end up being very short (i.e. <20 bases). These short reads are likely to align to multiple (wrong) locations on the reference, introducing noise into the variation calls. They can be removed using PrinSeq, Trimmomatic (using the MINLEN option), or a simple in-house script. Minimum acceptable read length should be chosen based on the length of sequencing fragment: longer for longer fragments, shorter for shorter ones – it is a matter of some experimentation with the data.

The three pre-processing steps above can be parallelized by chunking the initial fastq file (hundreds of millions of reads, up to 50-150 G of hard disk space per file depending on sequencing depth) into several files that can be processed simultaneously. The results can then be combined.

Bibliography