CONSeQuence is a program that attempts to predict highly detectable peptides for a given protein sequence. It combines
four machine learning algorithms; SVMlite, RandomForests, Artificial Neural Networks and Genetic Programming. Each algorithm was trained and tested on observed peptides (with up to 2 internal missed-cleavages) from Yeast experiments in Peptide Atlas.
The output produces peptides, of length between 6 and 42 residues, which are predicted to be detectable by 1,2, 3 or all 4 algorithms (CONS1-4). In addition you can select the number of internal missed-cleavages.
To use the program select a fasta file or paste in your sequence, select the number of internal missed-cleavages and the prediction type.
The prediction types available are:
Binary - Assigns peptides a score of 1-4 (the number of algorithms that predict to fly)
ANN only - Runs the Artificial Neural Networks only and provides a score between 0-1 (scaled raw outputs from ANN)
Rank score - Assigns peptides a ranking score between 0-1 (based on a linear SVM)
CONSeQuence accepts protein sequences in fasta format, with a limit of 5000 proteins per run. Running 5000 proteins with 2 miscleavages may take over 15 minutes. If you require whole proteome submission please email firstname.lastname@example.org.