NYCE subcellular location prediction

NYCE predicts subcellular location (either Nuclear, Nucleo-cytoplasmic, Cytoplasmic or Extracellular) of eukaryotic proteins using the predicted exposure value of their amino acids. Predictions with low scores (<0.4) can be taken as an indication that the sequence is in another location (e.g. Mitochondria).
Listed below are some frequently asked questions. If you have other questions or suggestions feel free to contact us.

* How to calculate residue exposure values from sequence only?
* How does NYCE work?
* I have many protein sequences. How do I predict localization?
* umm...I think there is a bug!!!

Calculating residue exposure values using the SABLE tool
Please note that the SABLE tool is created and maintained by the group of Prof. Jarek Meller.
Residue exposure values for a protein sequence can be calculated using the SABLE tool. Here is a step by step procedure:
1. Go to the SABLE website
2. Paste your sequence in the "Amino acid sequence" box and click on "Submit"
3. Wait for the result to appear. Typical running times are about a minute or two.
4. In the result page (POLYVIEW-2D) go to the "Get sequence" section. Check the box "AA" and "RSA"
5. It will open the sequence and relative solvent accessibility in a new window
6. Copy and paste the sequence and relative solvent accessibility in the NYCE form

How does NYCE work?
From the residue accessibility values NYCE calculates the frequency distribution of amino acids at different range vales and makes 3 different range vectors. Each range vector is fed to 4 different one-vs.-rest SVM models (one for each location class). The probability values obtained from these SVM models are combined by an optimized ANN model. The ANN gives a score for each of the 4 location classes. For more details see the reference.

How does NYCE work?

From the residue accessibility values NYCE calculates the frequency distribution of amino acids at different range vales and makes 3 different range vectors. Each range vector is fed to 4 different one-vs.-rest SVM models (one for each location class). The probability values obtained from these SVM models are combined by an optimized ANN model. The ANN gives a score for each of the 4 location classes. For more details see the reference.

Prediction for multiple sequences
NYCE supports only one sequence at a time. For batch prediction please contact us.

Bug/Technical problem!!!
We try to make NYCE really nice. But if you observe any problem and/or have any suggestion feel free to contact us.

Reference