摘要： In silico genome analysis enables systematic identification
of potential antigens within a pathogen. Of primary importance is
the accuracy of computer algorithms used for antigen prediction.
Most bioinformatics tools are based on sequence alignment and are
not able to predict truly novel antigenic proteins which lack similarity
to existing antigens or which encode antigenicity in a cryptic manner.
To surmount such obstacles, we have recently developed an alignment-free
approach to in silico antigen identification, based on
the auto cross covariance (ACC) transformation of protein sequences
into uniform vectors of principal amino acid properties. Here, we
apply this approach to finding parasite and fungal immunoprotective
antigens. The models derived in this study demonstrate good predictive
ability with 78% to 97% accuracy under internal cross validation
in 7 groups. Under external validation, they gave 69% sensitivity
ranking the true immunoprotective proteins in the first 25% of their
proteomes.
