_PROBLEM CoEPrA-2006_Classification_001

_GROUP_NAME Alexander Zelikovsky

_GROUP_MEMBERS
Alexander Zelikovsky
Dumitru Brinza

_ADDRESS
Alexander Zelikovsky
Department Computer Science, 
Georgia State University
Phone: (404) 651-0676 
Fax: (815) 642-0052
Email: alexz@cs.gsu.edu
Office: 1443, Peachtree Str. 34
web:http://www.cs.gsu.edu/~cscazz/

Dumitru Brinza
Department Computer Science, 
Georgia State University
Phone: (404) 463-2808
Email: dima@cs.gsu.edu
Office: 1415, Peachtree Str. 34
web:http://www.cs.gsu.edu/~cscdubx/

_MODELING_PROCEDURE

METHOD DESCRIPTION

The following approach was succesfuly used in disease susceptibility
prediction based on genotype data. Bellow you can find a link with
full description of our method.

http://suez.cs.gsu.edu/%7Ecscazz/postscript/wabi06.pdf

For CoEPrA competition we slightly modify our method to handle amino
acid's alphabet in place of 0,1,2 which we have used for SNP notations.

Short description of the proposed method:

1. For each protein sequence s we find the most significant multi-amino-acid 
   combination which attributes s to class +1 (MAAC+) and combination which
   attributes s to class -1 (MAAC-). These combinations are found using 
   exhaustive search on the training (calibration) dataset. For datasets with
   longer protein sequencies we use combinatorial and greedy searches, which
   are described in http://suez.cs.gsu.edu/%7Ecscazz/postscript/bgrs06.pdf
2. Based on the significance (p-value) of MAAC+ (SG+) and significance of MAAC- 
   (SG-) we decide if s belongs to the class +1 or -1. If (SG+)/(SG-) < CONST 
   then s belongs to -1 else to +1. The value of CONST is computed as follows.
3. On the training dataset we perform leave-one-out test by computing for each 
   left out sequence CC = (SG+)/(SG-). Then we sort all sequencies by CC and
   find a position such that all sequencies below are classified to -1 and 
   above to +1. Position is chosen to minimize the error. 

METHOD VALIDATION

We validated our method by performing leave-one-out test on training (calibration)
data CoEPrA-2006_Classification_001_Calibration_Peptides.txt. The result of
the running can be found below. The estimated accuracy is 76.4%.

_PREDICTION
Obj_00001 +1
Obj_00002 +1
Obj_00003 +1
Obj_00004 -1
Obj_00005 -1
Obj_00006 -1
Obj_00007 -1
Obj_00008 -1
Obj_00009 -1
Obj_00010 -1
Obj_00011 -1
Obj_00012 -1
Obj_00013 -1
Obj_00014 -1
Obj_00015 -1
Obj_00016 -1
Obj_00017 -1
Obj_00018 -1
Obj_00019 +1
Obj_00020 -1
Obj_00021 +1
Obj_00022 -1
Obj_00023 -1
Obj_00024 +1
Obj_00025 -1
Obj_00026 +1
Obj_00027 -1
Obj_00028 +1
Obj_00029 +1
Obj_00030 -1
Obj_00031 -1
Obj_00032 +1
Obj_00033 +1
Obj_00034 -1
Obj_00035 -1
Obj_00036 -1
Obj_00037 -1
Obj_00038 +1
Obj_00039 -1
Obj_00040 +1
Obj_00041 +1
Obj_00042 +1
Obj_00043 -1
Obj_00044 +1
Obj_00045 +1
Obj_00046 +1
Obj_00047 +1
Obj_00048 -1
Obj_00049 +1
Obj_00050 +1
Obj_00051 +1
Obj_00052 +1
Obj_00053 -1
Obj_00054 +1
Obj_00055 +1
Obj_00056 +1
Obj_00057 +1
Obj_00058 +1
Obj_00059 +1
Obj_00060 +1
Obj_00061 +1
Obj_00062 +1
Obj_00063 +1
Obj_00064 -1
Obj_00065 +1
Obj_00066 +1
Obj_00067 +1
Obj_00068 -1
Obj_00069 +1
Obj_00070 +1
Obj_00071 +1
Obj_00072 +1
Obj_00073 -1
Obj_00074 +1
Obj_00075 +1
Obj_00076 +1
Obj_00077 +1
Obj_00078 +1
Obj_00079 +1
Obj_00080 +1
Obj_00081 +1
Obj_00082 +1
Obj_00083 +1
Obj_00084 +1
Obj_00085 +1
Obj_00086 +1
Obj_00087 +1
Obj_00088 +1