We have a big dataset of mtDNA COI sequences, around 1500 with around 30 insect species. We used DAMBE (http://dambe.bio.uottawa.ca/dambe.asp) to condense sequences into haplotypes. However, after alignment with clustalX and trimming with Gblocks (http://molevol.cmima.csic.es/castresana/Gblocks_server.html) of all the poorly aligned regions, we managed to get a compression of only ~ 50% (800 haplotypes). Many sequences have a very small degree of divergence, sometimes less than 1%. I'd like to know if you know a program capable of condensing slightly divergent sequences into consensus sequences. Another strategy we would follow would be to build a phylogenetic tree and select sequences which cluster together with small branch length between each other, but it would be certainly more time-consuming.
Thank you in advance for your help!
Dear Pierfrancesco,
May be you should try DIVEIN which is a web interface allowing reconstruction of the consensus, most recent common ancestor, and center of tree sequences. DIVEIN also provides tools for further analyses, including condensing sequence alignments to show only informative sites or private mutations.
You can download the software from this site:
http://indra.mullins.microbiol.washington.edu/DIVEIN
Good luck !
Hello Pierfrancesco,
What program did you use to edit your sequences before alignment and condensing your haplotypes? If you use Sequencer you can create contigs by a given clustering thrush hold.
Using a tree approach should work if you know that you will have clearly defined OTUs under a given sequence divergence. If what you are trying to do is just break them up into OTUs given a certain sequence divergence you could use Uclust.
http://drive5.com/usearch/manual/uclust_algo.html
Also a word to the wise, make sure you have confidence in you sequence calls, I have seen a lot of people have weird results and waste a lot of time due to not editing the sequences or not trimming the sequence based on a quality score, depending on your sequencing method.
Best Regards
Aloha Pierfrancesco,
Yes the point near the end of the previous answer by Max Bangs was exactly my concern, i.e. the stretches of sequence data at the ends of read are usually less accurate and so some of the differences might be due to misreads. The presence of 800 haplotypes is...unusual! One method is to try some slightly more extreme levels of trimming, until you reach a point where the number of haplotypes is more manageable. If you can retain the hot spots that you have confidence in, i.e. they are the real mutations/differences, then you're in business.
Thank you very much for your contributions. QC was definitely an issue. Unfortunately it is a big dataset and not my main job, so I couldn't QC again all the original .ABIs. I have solved the problem now; on the one hand, with more extreme trimming as you suggested; on the other hand, splitting the dataset. I divided the dataset in subsets of genera (according to blast results). Then I aligned and trimmed ruthlessly! Then, I used the ALTER online service (http://sing.ei.uvigo.es/ALTER/) to compress the sequences into haplotypes and it worked perfectly. DAMBE still refuses to compress them into haplotypes though. Perhaps it is some malfunction related to the mac version of the program?
BTW thank you again!
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools