pseapred

'pseapred' is a software tool developed to predict secretory proteins in malaria parasites. These secretory proteins are of particular interest because they play a crucial role in the survival and growth of the malaria parasite within infected red blood cells (RBC). Identifying these secretory proteins is essential for the development of vaccines and drugs to combat malaria. Traditional motif-based methods for predicting secretory proteins have had limited success due to the lack of a universal motif that applies to all secretory proteins of the malaria parasite.
To address this challenge, 'pseapred' takes a systematic approach to develop a general method for secretory protein prediction. The tool employs machine learning techniques, specifically Support Vector Machines (SVM), to predict secretory proteins based on their amino acid composition, dipeptide composition, and other features. The software was trained and tested on a non-redundant dataset consisting of 252 secretory proteins and 252 non-secretory proteins.
Several SVM models were developed, and they achieved high accuracy and Matthews Correlation Coefficient (MCC) values. For instance, the model using amino acid composition achieved an accuracy of 85.65% and an MCC of 0.72, while the model based on dipeptide composition reached an accuracy of 86.45% and an MCC of 0.74. Additionally, models using split-amino acid and split-dipeptide composition were developed and achieved even better results, with an MCC of 0.74 and an accuracy of 86.40% for the split-amino acid model and an MCC of 0.77 and an accuracy of 88.22% for the split-dipeptide model.
A significant innovation of 'pseapred' is the use of Position-Specific Scoring Matrix (PSSM) profiles obtained from PSI-BLAST, which has not been previously applied in secretory protein prediction for malaria parasites. Using PSSM profiles, the tool achieved the highest accuracy, with an MCC of 0.86 and an accuracy of 92.66%.