About SeqNLS

SeqNLS is a novel method for nuclear localization signal prediction based on frequent pattern mining and linear-motif scoring. SeqNLS can effectively detect potential NLS candidates extensively with consistant performance.

Usage


Input
Protein sequence(s) in fasta format, and final-score cutoff.
The higher selected final-score cutoff, the higher precision but lower recall for the prediction performance.

Output
The positions of NLS predictions within the query sequences.

Prediction result (The predicted NLS(s) are underlined)
YGR027C: MPPKQQLSKAAKAAAALAGGKKSKKKWSKKSMKDRAQHAVILDQEKYDRILKEVPTYRYVSVSVLVDRLKIGGSLARIALRHLEKEGIIKPISKHSKQAIYTRATASE
The matches with scores in different ranges are marked in different colors.

Definition of different colors in predictions
score rangecolor
0.1-0.3
0.3-0.5
0.5-0.7
0.7-0.8
0.8-0.86
0.86-0.89
>0.89


The predicted NLS(s) (score cutoff = 0.1)
Protein IDPredicted NLSStartStopThe highest score of matches within the prediction
YGR027CAKAAAALAGGKKSKKKWSKKSMKDRA11360.878


The estimated prediction performance of SeqNLS on this server
Score CutoffPrecisionRecallResidue Level Accuracy
0.10.22-0.30.8-0.810.43-0.55
0.30.3-0.40.7-0.780.47-0.58
0.50.37-0.530.64-0.710.5-0.59
0.70.49-0.590.56-0.660.57-0.64
0.80.55-0.650.56-0.630.58-0.65
0.860.67-0.720.44-0.560.6-0.66
0.890.72-0.830.18-0.380.62-0.8
Note: In our paper, PrDOS(http://prdos.hgc.jp/cgi-bin/top.cgi) is used as the disorder predictor; in this server, IUPred(http://iupred.enzim.hu/) is used as the disorder predictor since PrDOS has no standalone package available. In addition, the server version doesn't implement relatively local conservation (IRLC) based masking because it is very computationally intensive. The table above is the estimated prediction performance of SeqNLS implemented on this server.