I have >1500 sequence clusters, all have 7 orthologous sequences from different species. When I look at multiple sequence alignments of some of these sequence clusters, I found some of the sequences weakly aligned with the other sequence in the same cluster. I want to remove these sequences. Is there any tool available to remove such sequences from all clusters in an automated way.
Gblocks will just remove a badly aligned column across the entire sequences. Gblocks is also too conservative even genuinely diverging positions will just be treated as badly aligned position and will just be deleted. In an extreme cases you might remain with just less than 50 percent of your sequence with gblocks. In my opinion Gblocks wont help you much but this depends on your research.
Normally what I would do is to collect the right orthologs in the first place. By right orthologs I mean selecting orthologs from reasonably diverged species, ensuring that the sequences I keep do not differ much by length(gaps minimized). If you align highly divergent species(e.g. human vs mosquito sequence) you're not guaranteed to have straightforward alignments.
Doing preliminary data cleaning will help you serve a lot of your valuable time downstream. I guess right now going cluster after cluster, checking a sequence that introduced alignment problems, and deleting that sequence, is way too manual and very time consuming.
Hello Dear Ashmit,
Yes there is a tool "BioEdit" . You can remove your badly aligned sequence by deleting the bad alignments and improve your alignment. This software tool is freely available online for various operating system. There are also various versions of it present on web- http://www.mbio.ncsu.edu/bioedit/bioedit.html.
Hope it will help you....
There are various tools available in the database. I prefer CLC workbench. It is good and easy.
As far as automatic deletion is concern, I am not sure whether there are any software's available for deleting the bad alignments or sequences automatically.
You have to go through each and every sequences very carefully to get desired result.
All the best....!!!
Ashish
G-blocks will work a treat here: http://molevol.cmima.csic.es/castresana/Gblocks.html
This is new and may help...
http://www.biomedcentral.com/1756-0500/7/806
Gblock is recommended to used.
http://molevol.cmima.csic.es/castresana/Gblocks_server.html
Gblocks will just remove a badly aligned column across the entire sequences. Gblocks is also too conservative even genuinely diverging positions will just be treated as badly aligned position and will just be deleted. In an extreme cases you might remain with just less than 50 percent of your sequence with gblocks. In my opinion Gblocks wont help you much but this depends on your research.
Normally what I would do is to collect the right orthologs in the first place. By right orthologs I mean selecting orthologs from reasonably diverged species, ensuring that the sequences I keep do not differ much by length(gaps minimized). If you align highly divergent species(e.g. human vs mosquito sequence) you're not guaranteed to have straightforward alignments.
Doing preliminary data cleaning will help you serve a lot of your valuable time downstream. I guess right now going cluster after cluster, checking a sequence that introduced alignment problems, and deleting that sequence, is way too manual and very time consuming.
If I have that problem, I pre-select what to aligned based on several factors (length of alignment, e-value, identity, number of mismatches) or a combination of them. This can be pipelined using BLAST. Let me know, I can help you with a script if you need it.
Jorge
Ashmit, could you send me 10 good aligned sequences and 5 bad aligned sequences to use real data for the script?. I will write the script and post it so you can download it and run it with the rest of your data.
The GUIDANCE software and server will do the job (see http://guidance.tau.ac.il/, and corresponding paper http://www.ncbi.nlm.nih.gov/pubmed/20207713). It assesses the uncertainty in the alignment both sequence-wise and column-wise, enabling filtering of the alignment. You might also consider TrimAL, which has an option for filtering sequences and will most likely be faster for large data sets (http://trimal.cgenomics.org/ and http://www.ncbi.nlm.nih.gov/pubmed/19505945 for the reference).
Hi Ashmit,
As Julien mentioned, you can use trimAl for removing automatically the poor aligned sequences. The options "-resoverlap" and "-seqoverlap" should make the job. You can vary those parameters until you are happy with the results. On this way you can reproduce easily the results.
Apart of the mentioned links - you can get the latest trimAl version at https://github.com/scapella/trimal/.
And don't hesitate to contact us if you are not satisfied with the result - and have a better idea about how to trim your alignments. We always welcome feedback from our users and look forward to implement new methods.
You could also use Rsamtools and ShortRead packages in R Bioconductor to import the alignment and sort by the number of mismatches in the cigar field.
Hi Ashmit,
Here is the script. It is very simple that might work for you. It separate the sequences that aligns good or bad into separate files. You just need NCBI BLAST+. Just define the threshold based on percentage identity. Let me know if it does what you need to.
Here is the link: https://sourceforge.net/projects/blastfilter/
- Badges
- Science topic
- Similar topics
- Bioinformatics
- Bioinformatics and Computational Biology