Software Architecture
Original Approach:
Our first approach to this project was in designing a java program
that linked directly to NCBI's BLAST search engine and ran queries to that
site. Our software would then parse the data results returned from the BLAST
searches and input them in a readable form into our matrix algorithm. Our
matrix algorithm calculates distances relative to a source and target organism
and outputs the data graphically as a "tree". We chose to output
it graphically as a tree because our focus was ease-of-use and readability
for the end-user.
Modified Approach:
The problem with the original approach was the difficulty in
designing the web interface to interact with NCBI's BLAST site. Per suggestion
by Prof. Arkin, we've instead decided to create our own BLAST search locally
by downloading the necessary genome data from the internet and source code
for running BLAST.
Application Design:
Modular design divided into following portions:
(details of each are discussed in powerpoint presentation)
-
Design allows for reuse of components in different applications
with minor changes
-
Design allows individual subrountines to be used recursively
to generate desired results with minimal changes
Software Architecture
Input Description:
The input portion contains HTML/CGI files for user input and a Perl program
for running BLAST and parsing the result.
1. HTML/CGI part has two files: blast.htm and blast.cgi.
http://sahara.lbl.gov/~lyan/cgi-bin/blast.htm
The above web interface allows the user to input a pathway name, organism names,
protein names, multiple protein sequences, and a threshold. Blast.cgi will save
those information into 5 files: Pathwayname, org.file.txt, Proteinnames, Proteinsequences,
and Threshold. Then it will run the bp.pl9 program.
2. Protein Databases and BLAST tools (ie. blastall, fascacmd)
for 5 organisms are located in /usr2/people/lyan/public_html. Those Databases
and BLAST tools are used by "bp.pl9" which will do the following:
- Get the threshold value from the first command line argument
- Divide up the Proteinsequences file into individual sequence files and get
the name of the protein (for example, wg.seq) from "Proteinnames". These are
the 1st seed genes
- For each gene (ie. wg.seq), it will do the following:
- Run blastall against all organisms except drosophila, cut at threshold
and get 2nd seed genes, save result in *_out (eg, wg_out)
- Get the sequence of the 2nd seed genes using fastacmd
- Run those 2nd seed genes against all the other organisms except itself,
cut at the threshold and append all the results for every gene in *_out
- Run the next 1st seed genes one by one
- The final result will be several *_out files with names saved in gabe.txt
Output Display Description:
The output portion contains five classes in four different files. The classes
are as follows: a) intree, b) dphy, c) Orthotable, d) dismat2 and e) newtree.
They are stored in separate files except for dphy, which is saved as part of
newtree. Intree.java takes in a text file of the output of the clustering algorithm
and reads in the organisms, names and distances for drawing the phylogenetic
tree. Dphy.java creates a JPanel and contains the method to draw the phylogenetic
tree on the panel. Dismat2.java takes organism names and their distances from
Metric.java and paints a graphical display of the distance matrix in another
JPanel. Orthotable.java takes in organism names and a hashmap of vectors of
orthologs in the different organisms from Metric.java and counts the number
of orthologs to the seed genes in each organism. These are then displayed in
a JTable in a third JPanel. Newtree.java is the major display class which calls
methods from all the other classes and creates a tabPane containing the above
three JPanels inside a JFrame. Newtree is then called by Phylogenetics to display
the desired results.