Many people have noticed sequence content bias at the 5' end of Illumina RNA-Seq reads attributed to random hexamers used for priming. There has been a proposal to correct coverage for this bias, and some people clip the first 13 bases from 5' if their reads are long enough. Is clipping actually useful or does it just remove valuable sequence data?
Unfortunately, similar to whether you should remove low-quality reads off the 3' end of reads, I would suggest that it depends on if the tools you are using take this information into account.
I encourage you to read the papers and documentation of the tools you are using to determine if the bias is considered when processing, if the tool removes them automatically, or if information embedded in the bias is used to better analyze the data.
Unfortunately, I have no idea what the common practice for this is. You may do well looking at some of the methods sections of the papers from some of the prominent RNA-Seq universities. Or potentially contact a preprocessing data lab directly. Illumina likely keeps up with the common techniques used across RNA-Seq bioinformatics pipelines.
As Joseph says, the decision whether to use clipping rather depends on your application and the tools you are using.
However, I also see this bias at the 5' ends of reads from DNA libraries that were made by sonication, which I believe to be as close to random as it gets. So maybe its not all related to random hexamers?
I don't understand exactly how it happens... but I don't like it, so I often clip it off and sometimes it seems to make my de novo assemblies better (its iimportant to QC data carefully for that kind of work though). Also we use a MiSeq, so we get relatively long reads and don't mind trading some length for high quality.
I don't see it as a problem. The bias is minimal and you lose much more information from clipping good reads rather than retaining the potentially biased reads.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics