Thesis Help: Dna Sequence using BLAST

Question

[ismailkiron5]

Hello,

I am doing my undergrad research on an android application which will fetch scientific name and it's sequence with score using BLAST. we already implemented BLAST and needleman wunsch for Global and local alignment. But We need add some improved module from existing module. Can anyone suggest anything?

[samuelmiver]

Not sure if I understand you correctly. Do you need a module able to run remote BLAST? Which is the programming language you are using?

[pld]

I guess if you're using Android you could just use BioJava.

[samuelmiver]

Here is the implementation:

http://biojava.org/wiki/BioJava:CookBook3:NCBIQBlastService

[ismailkiron5]

I actually did the implementation from this

https://code.google.com/p/meta-proteome-analyzer/source/browse/branches/development/src/de/mpa/client/blast/RunMultiBlast.java?spec=svn769&r=769

[ismailkiron5]

I already did the implementation. The problem is I need show improved application from the existing one. It can be by adding new feature or new method.

[samuelmiver]

One of the critical things that can add minutes and minutes to a BLAST process is the word size. In order to find matches quickly, BLAST looks for perfect matches across subsequences of the length of this word size. Therefore, if the word size is very small, the search takes longer. If the search size becomes large, imperfect matches (e.g. with a single nucleotide insertion) are no longer found.

Depending in what you need performance vs. quality you can play with the word size to achieve a equilibrium between both of them.

[ismailkiron5]

Thank you so much for the reply. What are the things needed for adding minutes?

[samuelmiver]

I haven't worked with BLAST in Java but it is very universal the use of -W to set the word size. The default is 11 for blastn, 28 for megablast and 3 for all others. In case you want to improve the time I will try increasing a little bit those values.

This will reduce the performance time but could introduce artefactual results so I recommend you to take some sequences you previously know the correct result and test how much you can increase the word size without having errors.

[pld]

Is there already a sequence alignment tool for android?

[ismailkiron5]

no we are using the Java version in android as a tool

https://code.google.com/p/meta-proteome-analyzer/source/browse/branches/development/src/de/mpa/client/blast/RunMultiBlast.java?spec=svn769&r=769

Answer 1

[samuelmiver]

Just my previous comments together in an answer:

One of the critical things that can add minutes and minutes to a BLAST process is the word size. In order to find matches quickly, BLAST looks for perfect matches across subsequences of the length of this word size. Therefore, if the word size is very small, the search takes longer. If the search size becomes large, imperfect matches (e.g. with a single nucleotide insertion) are no longer found.

Depending in what you need performance vs. quality you can play with the word size to achieve a equilibrium between both of them.

I haven't worked with BLAST in Java but it is very universal the use of -W to set the word size. The default is 11 for blastn, 28 for megablast and 3 for all others. In case you want to improve the time I will try increasing a little bit those values.

This will reduce the performance time but could introduce artefactual results so I recommend you to take some sequences you previously know the correct result and test how much you can increase the word size without having errors.

[ismailkiron5]

I am using this algorithm . Can I merge it with the following algorithm

https://code.google.com/p/meta-proteome-analyzer/source/browse/branches/development/src/de/mpa/client/blast/RunMultiBlast.java?spec=svn769&r=769

[samuelmiver]

I am not familiarized with Java but if it works like bioperl for example, you should add the option -W and the word size selected to the line that calls BlastP from NCBI (in the same way you put the e-value for example).