[Biojava-l] Sequence Iteration in BioJava(x)

Fri Dec 16 00:37:30 EST 2005

You should also be aware that the biojavax sequence i/o is actually about 
3 times slower than the biojava sequence i/o (for genbank, haven't tested 
others). This is because it does a much better job of parsing the relevant 
details into a more structured object heirachy.

Having said that it is possible to set i/o pipeline up so that it ignores 
details that are not of interest to you. If you only want the sequence 
name and the sequence data from Genbank (and not all the features, 
annotations and comments) then parsing is on average about 10x faster 
(based on about 4000 eukaryote records). Details on how to do this can be 
found in the biojavax docboc in CVS.

- Mark

Mark Fortner <m.fortner at sbcglobal.net>
Sent by: biojava-l-bounces at portal.open-bio.org
12/16/2005 12:09 PM
Please respond to m.fortner

        To:     biojava-list <biojava-l at biojava.org>
        cc:     (bcc: Mark Schreiber/GP/Novartis)
        Subject:        Re: [Biojava-l] Sequence Iteration in BioJava(x)

Mark,
Thanks for the info.  This is sort of a test project for us.  We have a 
few classes and data structures in C++ that handle operations like 
sequence io and packing, and are fairly fast.  However, we've also come 
to the realization that we've spent a lot of time dealing with 
cross-platform and compiler-related problems, and if Java can give us 
comparable performance then we might switch to it.  If nothing else, the 
opportunity costs would be lower, since we could write and test more 
code, in the same amount of time.  The tools are good-deal better for 
Java development than C++.

We're at the point where we can either continue to invest time in our 
library or rewrite what we have using BioJava and other libraries.  I've 
written a lot of Java-code over the past 10 years and suggested that we 
try Java both using the standard javac compiler and gcj to see if we can 
get C++ like performance.

Thanks for your help,

Mark

mark.schreiber at novartis.com wrote:

>There is probably not any performance benefit except in the case of very 
>large sequences which are often compressed behind the scenes by biojava.
>
>The benefits may come from ease of use and object orientation.
>
>eg, There is probably already a parser to read in an validate your 
>sequence, The windowing or nMer stuff is already figured out and has been 

>used by lots of people so it's been "stress tested". Also the objects 
>themselves have a lot of functionality built in that a character stream 
>does not. The downside of using objects is they take up memory and there 
>is a certain amount of overhead in there construction. To help overcome 
>this SymbolLists are actually lists of references to Symbols not lists of 

>Symbols themselves. This makes them much smaller (although not as small 
as 
>char[]'s).
>
>If you want superfast performance then you should bit encode the data and 

>operate over it with memory pointers as in C or machine code. You should 
>be aware though that any intensive loop like the ones that would be used 
>to carry out this operation in biojava will almost certainly be detected 
>and compiled into native code by the Java Runtime on the fly. This might 
>make it hard to say if the java code would be much slower than the C 
code.
>
>- Mark
>
>
>
>
>
>Mark Fortner <m.fortner at sbcglobal.net>
>Sent by: biojava-l-bounces at portal.open-bio.org
>12/16/2005 10:36 AM
>Please respond to m.fortner
>
> 
>        To:     biojava-list <biojava-l at biojava.org>
>        cc:     (bcc: Mark Schreiber/GP/Novartis)
>        Subject:        Re: [Biojava-l] Sequence Iteration in BioJava(x)
>
>
>Richard,
>Thanks for the example.  Your approach is very similar to a non-BioJava 
>approach that I had worked out earlier.  I was wondering if the 
>BioJava(x) API provides any performance benefit over simply running a 
>window along a character stream? 
>
>The work that we're doing involves iterating through the human genome, 
>(and in a number of cases, metagenomic sequences) and we're trying to 
>squeeze as much performance out of it as possible while minimizing the 
>memory footprint.
>
>Thanks,
>
>Mark
>
>Richard HOLLAND wrote:
>
> 
>
>>orderNSymbolList splits the sequence into non-overlapping chunks. What
>>is required here is chunks that are only one base different (further on)
>>than the previous chunk.
>>
>>The simplest way would be this:
>>
>>               SymbolList mySeq; // this is your sequence from somewhere 

>> 
>>
>else
> 
>
>>               for (int i = 1 ; i <= mySeq.length()-2; i++) {
>>                               SymbolList trimer = mySeq.subSeq(i,i+2); 
>> 
>>
>// coords are
> 
>
>>inclusive so i to i+2 = 3 bases
>>                               // do something with your trimer here
>>               }
>>
>>Note that the index starts at 1 and goes right up to and including
>>length(), as symbols in a SymbolList are 1-indexed, not 0-indexed.
>>
>>cheers,
>>Richard
>>
>>Richard Holland
>>Bioinformatics Specialist
>>GIS extension 8199
>>---------------------------------------------
>>This email is confidential and may be privileged. If you are not the
>>intended recipient, please delete it and notify us immediately. Please
>>do not copy or use it for any purpose, or disclose its content to any
>>other person. Thank you.
>>---------------------------------------------
>>
>>
>>
>>
>> 
>>
>>>-----Original Message-----
>>>From: biojava-l-bounces at portal.open-bio.org 
>>>[mailto:biojava-l-bounces at portal.open-bio.org] On Behalf Of David Huen
>>>Sent: Friday, December 16, 2005 7:34 AM
>>>To: m.fortner at sbcglobal.net
>>>Cc: biojava-list
>>>Subject: Re: [Biojava-l] Sequence Iteration in BioJava(x)
>>>
>>>
>>>On Dec 15 2005, Mark Fortner wrote:
>>>I think what you want is the SymbolListViews.orderNSymbolList method.
>>>
>>>It will take a SymbolList and turn it into another where it 
>>>is viewed in 
>>>another compound alphabet of the required order.
>>>
>>>
>>>
>>>
>>> 
>>>
>>>>I'm looking for the best way to iterate through all
>>>>nmers within a given sequence.  For example, given a
>>>>sequence that looks like this:
>>>>
>>>>ACTGACTGACTG
>>>>
>>>>If I extract all trimers from this I should get:
>>>>
>>>>ACT
>>>>CTG
>>>>TGA
>>>>GAC
>>>>ACT
>>>>CTG
>>>>TGA
>>>>GAC
>>>>ACT
>>>>CTG
>>>>
>>>>Is there an existing class that will allow me to
>>>>iterate through a sequence this way, or am I on my
>>>>own?
>>>>
>>>>
>>>>
>>>> 
>>>>
>>>_______________________________________________
>>>Biojava-l mailing list  -  Biojava-l at biojava.org
>>>http://biojava.org/mailman/listinfo/biojava-l
>>>
>>>
>>>
>>> 
>>>
>>
>> 
>>
>
>_______________________________________________
>Biojava-l mailing list  -  Biojava-l at biojava.org
>http://biojava.org/mailman/listinfo/biojava-l
>
>
>
>
> 
>

_______________________________________________
Biojava-l mailing list  -  Biojava-l at biojava.org
http://biojava.org/mailman/listinfo/biojava-l