CRUSH - CoReceptor USage prediction for HIV-1
CRUSH is an SVM model for the prediction of HIV-1 coreceptor usage based on the amino acid sequence of
the V3-loop of envelope glycoprotein gp120. CRUSH utilizes inter-residue interaction energies derived
from computational models of coreceptor complexes (V3-loop:CCR5 and V3-loop:CXCR4) as features. Additionally,
four known rules for coreceptor usage, namely net charge, glycosylation motif, 11/24/25 rule, and length,
are also used as features.
CRUSH takes as input a list of V3-loop sequences in FASTA format (example shown below) containing the first and last cysteine residues, and only consisting of the 20 natural amino acids. The sequences should not contain any ambiguous positions and should not contain any special characters, since CRUSH will only return results for sequences meeting this criteria. CRUSH returns a table containing the sequence IDs (a generic one is generated if not provided) and the predicted probability of the sequence being X4-tropic. At present, the method has not been trained on dual-mixed tropic (R5X4) virsues. In initial testing, a probability threshold of 0.75 provided a false positive rate of 0.05, while a probability threshold of 0.90 provided a false positive rate of 0.02.
Submit Protein Amino Acid Sequence
Kieslich, C. A., Tamamis, P., Guzman, Y. A., Onel, M., and Floudas, C. A. Near perfect prediction of HIV-1 coreceptor usage reveals the interactions driving tropism. PLOS One, 11(2): e0148974, 2016.