Hi Ryan,<br><br>Pick a number of sequence pairings, some where you know that the sequences are related and some where you know that they are not related. Your matrices should hopefully provide better discrimination than ednafull or BLAST, that is, they should provide more positive scores for the true positives and more negative scores for the known negatives. You need to do this on a large number of sequences to get a good picture- one or two is not enough.
<br><br>Best,<br>Marty<br><br><br><div><span class="gmail_quote">On 4/11/06,
<b class="gmail_sendername">Ryan Golhar</b> <<a href="mailto:golharam@umdnj.edu" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">golharam@umdnj.edu</a>> wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
I've constructed a few scoring matrices for DNA sequence based on<br>certain information. I'm now trying to determine how well they perform<br>compared to the standard statistically based matrices, such as EDNAFULL<br>(+5/-4) and the standard scoring matrix used by BLAST (+1/-3).
<br><br>My question is, what would be a valid test to compare matrices against<br>each other?<br><br>Ryan<br><br>_______________________________________________<br>Bioinformatics.Org general forum - <a href="mailto:BiO_Bulletin_Board@bioinformatics.org" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">
BiO_Bulletin_Board@bioinformatics.org</a><br><a href="https://bioinformatics.org/mailman/listinfo/bio_bulletin_board" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">https://bioinformatics.org/mailman/listinfo/bio_bulletin_board
</a><br></blockquote></div><br><br clear="all">
<br>-- <br>-- <br>Martin Gollery<br>Associate Director<br>Center For Bioinformatics<br>University of Nevada at Reno<br>Dept. of Biochemistry / MS330<br>775-784-7042<br>-----------