We currently aim to sequence human DNA with 90x coverage on a NovaSeq6000 system (TruSeq DNA PCR-Free Library Prep). The run parameters are all fine and with around 1.2 Tb the output should be sufficiently large (3 samples per flow cell). However, when we take a closer look at the coverage of our samples we can see that some regions of the genome are covered with up to 300-400x whereas a broad range of regions is covered with less than 40x (graphics attached). At the same time we have a very small percentage of duplicates (~6%). We are wondering where the selective amplification comes from?
Some details and stats for the alignment:
- Aligner: Isaac (Illumina, iSAAC-03.16.02.19)
- Reference genome: Homo sapiens (Ensembl GRCh37)
- Total aligned reads: ~2,000,000,000 (~93%)
- Fragment length: ~394bp
- Percent duplicate proper read pairs: ~6%
Many thanks in advance!
I think that there is nothing wrong with the post-run analysis (alignment). My first thought is that there is an issue with the fragmentation (unlikely if you use covaris) or end-repair and/or adaptor ligation step where not all fragments are successfully ligated with adaptors (non-compatible fragment ends ---> failed end-repair, DNA contaminats --> salts/inhibitors or poor quality of reagents --> (almost) dead ligase enzyme). You can re-purify gDNA, prolong ligation time and/or change the reagents lot. Hope it helps... Regards,J
I wonder how you assessed the percent duplication, since you would (in theory) not expect those. Usually duplication detection is done by checking whether reads or fragments have same sequence and the same start and end coordinates. This can also be true for random fragments, so probably marking PCR duplicates makes no sense here.
Besides, coverage values can be affected by several things. PCR is the obvious one, but shouldn't play a role here. On the NovaSeq sequencing happens within the structured flowcell by exclusion amplification, so some aspects of PCR based libraries are probably still retained.
And then there is mappability: Repeats and other non-unique sequences have a great influence on coverage values.
Still, the bimodal distribution of your coverage values looks somehow skewed. All of this can neither be explained by template-exclusion amplification.
So just wild speculation from here on: Since you work at a leukemia laboratory: Does this sequence come from a cancer genome. If so: copy-number variations have a drastic effect on coverage, if you expect something like a triploid tumor with several chromosomes being present in only one or two copies, that could look similar
Thanks to both of you for your answers. I also agree that the alignment itself will most likely not be the underlying reason for the uneven coverage that we observe for some samples. Since we do have examples of a very even coverage and all samples are processed in the same way(+ we already sequenced many samples). Unfortunately we haven't found any correlation that would give us a hint at the underlying problem. Fragmentation is done by covaris and reagent lots do change over time (no correlation with coverage distribution). We have changed different parameters of the workflow but couldn't find the answer yet.
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Biology
- Biomolecules
- Nucleic Acids
- DNA