I had 2 runs from the same library at some NGS platform (intentionally not mentioning the name of the platform) and I have ended up with some strange findings. In my second run I found some extra variable regions which were not there in the first.
FYI: I have used the same parameters in the analysis pipeline for both sets of data. Quality scores is above Q30 for all those alternative alleles in the second run. Locus been not reported as variant site before.
What could be the reason behind such a result?
Do you have the same coverage (amount of reads) for the two platforms? Maybe you are detecting a variable region that is present in a small number of cells.
The platform used may be very pertinent to your anomalous results...each has their own foibles. It is hard to comment without knowing the origin of the data ie, transcriptome, genome, platform used..
Are you referring about raw data on short reads (average lenght =50) or long reads (average 300) or you consider reads that have been assembled?
Have your samples been multiplexed and in that case were the same barcodes used?
What is the type of your NGS data (454, illumina...). Are the extra variable regions obtained on the reads or on the mapping data/de novo?
Different platforms? If so, this is not so surprising. If not, there is another reason, regions with low coverage always varies. To confirm this, you can check whether these variable regions overrepresented low-coverage regions.
The lack of experimental details seems to be a problem for most post I visit to try to offer advice. There needs to be a policy or standard set of information that is offered up to get proper advice. Check out some of the QA spaces in the computer science realms to see what I mean (eg. stackoverflow). I sort of understand most peoples reluctance to provide details as they fear being scooped, but if your want help you have to give a sufficient amount of detail, most problems stem from experimental artifacts or fundamental misunderstanding of biological principals. For example you can hind the name of a gene but should provide functional details like if it is a transcription factor or a kinase, I think the name of the sequencing platform or details on how a library is prepared are very pertinent pieces of information that will not deluge to the world your research.
The post on this system are too much like twitter. Its fine for quips or to point people at a resource/news article/event, but not okay for any serious conversation.
If the same material for the library went through two separate emPCR steps, one may expect a difference. If this was not the case, but there was a time lapse between the two runs greater than ten days, this may also contribute to the difference.
Difficult to discuss this problemwith limited information provided. How long have you stored the library after the first run? Are you talking of known variants or novel? You may compare between runs if you have validated the variants as true variants with a different platform (e.g. SNP genotyping). Then, check run1 vs run2 and run1 vs run1+2 to assess if limited coverage is an issue.
What does "some" difference mean? How big of a problem is this, have you done a correlative analysis? There will always be some difference. This is a stochastic process at a theoretical saturation point. Lilia has given you two excellent points in the process subject to this but there are many others.
Navonil,
can you maybe give details about the type of experiment? ChIP-seq are very different (in the biology) from RNA-seq and much more sensitive to replicate differences. I think this is the minimum of information required to give a qualified answer ...
How was the variant analysis done? Depending on the software, the quality characteristics of an individual run may have large impact on the fraction of called variants. I have experienced this with a combination of Illumina RNA-seq and soapsnp.
btw: Do you really mean a variable "region" (consecutive sites, not just a variable site? Note that NGS sequencing typically comes with context-specific error signatures. A G+C rich region, for example, is very error-prone on the Illumina platform. Depending on the coverage (and overall quality) of a sequencing run, a G+C rich region may show up as variable in one run but not in another.
Yes. Sorry I was out of network for long...1st ly both of them were IonAmpliseq data. Average coverage of the run was 150X. The particular site (giving data for one single such site was 285 and 315. The second variant occured in the second case only. The variants are not reported before. second variant occured in 22% of the total 315 reads. Days between two runs was 12. Yes template preparation was different ( in 2 spell of 12 days difference).
All of them are point variation,32 in number. Total coverage was 2kb
Contamination?
Once we found bird genes in cat sequencing and have been wondering what was the cat's diet :):) Then it turned out to be conserved ones from yet another species. Blast/map the irregularities against other references.
I don't want to consider it now... Since we were very cautious about these issues from the very beginning.At the end, If I really do not have any proper explanation, then may be have to consider it as contamination.
By the way, Lilia, how frequently did you observed such differences? Because, this happened only with one chip run. In the second chip, I got nothing so significant difference ( @ Here Significant does not mean statistical significance)
.
Navonil,
We see differences very often; however we work with viral quasispecies so this is to be expected. We had to develop filtering and alignment algorithms that remove the “background” noise based on sequencing of biological clones. I cannot claim that those algorithms are universal, though. I believe them to be very much target-specific.
I will be travelling all through the day... So may be I can'n answer any of your query today. Thank you for all valuable suggessions.
It sounds like there is more technical variability than you expected. However, it is quite likely that biological and unavoidable variations in library preparation will be substantially larger that technically variations. For a reproducible results is must be the case that "Reproducibility also refers to the ability of an entire experiment or study to be reproduced, or by someone else working independently. It is one of the main principles of the scientific method. The result values are said to be commensurate if they are obtained (in distinct experimental trials) according to the same reproducible experimental description and procedure." http://en.wikipedia.org/wiki/Reproducibility. Thus using only one library it is not possible to determine the reproducibility of your results, because you have no information on inter-replicate variability.
There are numerous reasons for large genomic variations in samples.. for instance tumor cells are known for genetic instability.. other examples are mosaicism.. viral and transposon elements.. some cells like lymphocytes go through processes of recombination such as those involved in antibody production.. the first step you do when you find an unknown sequence is to do a BLAST search and try to identify its origins.. all these are based on the presumption that there were no problem in data analysis
318 it is. And its not any cancer panel. And even if so, why do you think there would be variability?
- Badges
- Science method