Novel machine learning method allerStat identifies statistically significant allergen-specific patterns in protein sequences

pLM4Alg: Protein Language Model-Based Predictors for Allergenic Proteins and Peptides

“…ACC of our models ranged from 91.2% to 92.8%, higher than those of AllerTOPv2 (88.7%) and of AllergenFP (87.9%), using the same data set. Furthermore, our models had an AUC from 96.8% to 97.7%, comparing to that (87.8%) of the AllerStat model. ,, These results highlight the superiority of pLM-based models over the traditional modeling methods in allergenicity prediction.…”

“…Furthermore, our models had an AUC from 96.8% to 97.7%, comparing to that (87.8%) of the AllerStat model. 19,26,28 These results highlight the superiority of pLMbased models over the traditional modeling methods in allergenicity prediction.…”

“…Traditional ML methods (e.g., random forest, support vector machine, k-nearest neighboring, Naïve Bayes, etc.) have been the mainstream approaches in previous studies on allergenicity prediction. ,,,,,,, Deep learning methods generally have higher model complexity than traditional ML modeling methods. As allergy research gains increasing attention, it is anticipated that the data set sizes will continuously grow in the future.…”

“…Subsequently, various approaches are proposed from different perspectives for allergen prediction, such as motif-based approaches, similarity searches, machine learning-based modeling, etc. Prediction tools, such as AllerCatPro 2.0, AlgPred 2.0, have been used extensively for preliminary allergenic risk assessment and identification of new allergens. ,,− One representative project derived from the FAO/WHO guidelines was AllerCatPro/AllerCatPro 2.0, which implemented a hierarchical workflow employing five criteria with improved similarity-checking methods for both sequences and 3D structures. , The criteria used in AllerCatPro were based on statistical analysis of existing allergens, representing a form of experience/knowledge-based decision-making. For instance, if a protein sequence shares >35% identity with known allergens over 90 windows, it would be classified into the group with strong evidence for allergenic potential. , So far, this approach has achieved a great performance in the allergen benchmark data sets .…”

“…Another representative technique is the machine learning (ML) approach where the classification criterion is determined by the ML model by learning from a separate training data set. ,,,− , Besides the selection of modeling method, the most challenging task in the ML approach is the representation or encoding of protein/peptide sequences into a numerical vector/matrix. Various encoding methods, including amino acid descriptors, amino acid composition (AAC), pseudoamino acid composition (PseAAC), dipeptide composition (DPC), amino acid descriptors (AAD), position-specific scoring matrix (PSSM), physicochemical descriptors, biomedical properties, k-mer dictionary-based binary representation, etc., have been widely used in predicting allergenicity and other properties/bioactivities. ,,,,,− However, these features may not always accurately represent protein sequences and simple combinations can cause high-dimensional problems as well as the feature redundancy …”