Genome-wide identification of specific oligonucleotides using artificial neural network and computational genomic analysis

Table 4 The integration of ANN and BLAST (IAB algorithm).

IAB^a (T,N,F)^b
1.	size ← length [T ] - N
2.	Oligo allOligo [size]/* the data structure Oligo includes score and sequence*/
3.	*for* pos ← 1 to size
4.	do allOligo [pos].sequence ← T.substring (pos, pos + N)
5.	/ calculate the ANN score for each sliding oligo using the trained ANN /
6.	allOligo [pos].score = CalScoreByNN ^c (allOligo [pos ])
7.	sort the allOligo array into non-decreasing order by ANN score
8.	lowestSim ← 1.0
9.	*for* p ← 0 to(size • F^d)
10.	do oligo ← allOligo [p].sequence
11.	similarity ← CalSimByBlast ^e (oligo)
12.	if similarity < lowestSim
13.	*then* bestOligo ← oligo
14.	lowestSim ← similarity
15.	if lowestSim <TH_sim^f
16.	*then* *return* bestOligo
17.	*return* bestOligo

^aIAB, integration of artificial neural network (ANN) and basic local alignment search tool (BLAST). ^bThe input parameters are as follows: the THC sequence T, the oligo length N, and the sensitivity factor F; ^cCalScoreByNN procedure calculates the ANN score (indicates the cross homology) by trained ANN; ^dThe sensitivity factor is set as 0.2, 0.3, and 0.4; ^eCalSimByBlast procedure calculates the cross homology by WU-BLAST; ^fThe cross homology threshold (TH_sim) used for 70-mer and 50-mer was 0.7 and that used for 25-mer was 0.8.

ISSN: 1471-2105