Prediction and analysis of antifreeze proteins

VotePLMs-AFP: Identification of antifreeze proteins using transformer-embedding features and ensemble learning

Antifreeze proteins (AFPs) are a unique class of biomolecules capable of protecting other proteins, cell membranes, and cellular structures within organisms from damage caused by freezing conditions. Given the significance of AFPs in various domains such as biotechnology, agriculture, and medicine, several machine learning methods have been developed to identify AFPs. However, due to the complexity and diversity of AFPs, the predictive performance of existing methods is limited. Therefore, there is an urgent need to develop an efficient and rapid computational method for accurately predicting AFPs. In this study, we proposed a novel predictor based on transformer-embedding features and ensemble learning for the identification of AFPs, termed VotePLMs-AFP. Firstly, three types of feature descriptors were extracted from pre-trained protein language models (PLMs) during the feature extraction process. Subsequently, we analyzed six combinations generated by these three embeddings to explore the optimal feature set, which was input into the soft voting-based ensemble learning classifier for the identification of AFPs. Finally, we evaluated the model on the two benchmark datasets. The experimental results show that our model achieves high prediction accuracy in 10-fold cross-validation (CV) and independent set testing, outperforming existing state-of-the-art methods. Therefore, our model could serve as an effective tool for predicting AFPs.

A stacking-based algorithm for antifreeze protein identification using combined physicochemical, pseudo amino acid composition, and reduction property features

Antifreeze proteins have wide applications in the medical and food industries. In this study, we propose a stacking-based classifier that can effectively identify antifreeze proteins. Initially, feature extraction was performed in three aspects: reduction properties, scalable pseudo amino acid composition, and physicochemical properties. A hybrid feature set comprised of the combined information from these three categories was obtained. Subsequently, we trained the training set based on LightGBM, XGBoost, and RandomForest algorithms, and the training outcomes were passed to the Logistic algorithm for matching, thereby establishing a stacking algorithm. The proposed algorithm was tested on the test set and an independent validation set. Experimental data indicates that the algorithm achieved a recognition accuracy of 98.3 %, and an accuracy of 98.5 % on the validation set. Lastly, we analyzed the reasons why numerical features achieved high recognition capabilities from multiple aspects. Data dimensionality reduction and the analysis from two-dimensional and three-dimensional views revealed separability between positive and negative samples, and the protein three-dimensional structure further demonstrated significant differences in related features between the two samples. Analysis of the classifier revealed that Hr*Hr, HrHr, and Sc-PseAAC_1, 188D(152,116,57,183) were among the seven most important numerical features affecting algorithm recognition. For Hr*Hr and HrHr, supportive sequence level evidence for the reduction dictionary was found in terms of conservation area analysis, multiple sequence alignment, and amino acid conservative substitution. Moreover, the importance of the reduction dictionary was recognized through a comparative analysis of importance before and after the reduction, realizing the effectiveness of the dictionary in improving feature importance. A decision tree model has been utilized to discern the distinctions between dipeptides associated with the physical and chemical properties of His(H), Iso(I), Leu(L), and Lys(K) and other dipeptides. We finally analyzed the other seven features of importance, and data analysis confirmed that hydrophobicity, secondary structure, charge properties, van der Waals forces, and solvent accessibility are also factors affecting the antifreeze capability of proteins.

Accelerating Antimicrobial Peptide Discovery for WHO Priority Pathogens through Predictive and Interpretable Machine Learning Models

The escalating menace of multidrug-resistant (MDR) pathogens necessitates a paradigm shift from conventional antibiotics to innovative alternatives. Antimicrobial peptides (AMPs) emerge as a compelling contender in this arena. Employing in silico methodologies, we can usher in a new era of AMP discovery, streamlining the identification process from vast candidate sequences, thereby optimizing laboratory screening expenditures. Here, we unveil cutting-edge machine learning (ML) models that are both predictive and interpretable, tailored for the identification of potent AMPs targeting World Health Organization's (WHO) high-priority pathogens. Furthermore, we have developed ML models that consider the hemolysis of human erythrocytes, emphasizing their therapeutic potential. Anchored in the nuanced physical-chemical attributes gleaned from the three-dimensional (3D) helical conformations of AMPs, our optimized models have demonstrated commendable performance-boasting an accuracy exceeding 75% when evaluated against both low-sequence-identified peptides and recently unveiled AMPs. As a testament to their efficacy, we deployed these models to prioritize peptide sequences stemming from PEM-2 and subsequently probed the bioactivity of our algorithm-predicted peptides vis-à-vis WHO's priority pathogens. Intriguingly, several of these new AMPs outperformed the native PEM-2 in their antimicrobial prowess, thereby underscoring the robustness of our modeling approach. To elucidate ML model outcomes, we probe via Shapley Additive exPlanations (SHAP) values, uncovering intricate mechanisms guiding diverse actions against bacteria. Our state-of-the-art predictive models expedite the design of new AMPs, offering a robust countermeasure to antibiotic resistance. Our prediction tool is available to the public at https://ai-meta.chem.ncu.edu.tw/amp-meta.

Accurate Prediction of Antifreeze Protein from Sequences through Natural Language Text Processing and Interpretable Machine Learning Approaches

Antifreeze proteins (AFPs) bind to growing iceplanes owing to their structural complementarity nature, thereby inhibiting the ice-crystal growth by thermal hysteresis. Classification of AFPs from sequence is a difficult task due to their low sequence similarity, and therefore, the usual sequence similarity algorithms, like Blast and PSI-Blast, are not efficient. Here, a method combining n-gram feature vectors and machine learning models to accelerate the identification of potential AFPs from sequences is proposed. All these n-gram features are extracted from the K-mer counting method. The comparative analysis reveals that, among different machine learning models, Xgboost outperforms others in predicting AFPs from sequence when penta-mers are used as a feature vector. When tested on an independent dataset, our method performed better compared to other existing ones with sensitivity of 97.50%, recall of 98.30%, and f1 score of 99.10%. Further, we used the SHAP method, which provides important insight into the functional activity of AFPs.

AFP-SPTS: An Accurate Prediction of Antifreeze Proteins Using Sequential and Pseudo-Tri-Slicing Evolutionary Features with an Extremely Randomized Tree

The development of intracellular ice in the bodies of cold-blooded living organisms may cause them to die. These species yield antifreeze proteins (AFPs) to live in subzero temperature environments. Additionally, AFPs are implemented in biotechnological, industrial, agricultural, and medical fields. Machine learning-based predictors were presented for AFP identification. However, more accurate predictors are still highly desirable for boosting the AFP prediction. This work presents a novel approach, named AFP-SPTS, for the correct prediction of AFPs. We explored the discriminative features with four schemes, namely, dipeptide deviation from the expected mean (DDE), reduced amino acid alphabet (RAAA), grouped dipeptide composition (GDPC), and a novel representative method, called pseudo-position-specific scoring matrix tri-slicing (PseTS-PSSM). Considering the advantages of ensemble learning strategy, we fused each feature vector into different combinations and trained the models with five machine learning algorithms, i.e., multilayer perceptron (MLP), extremely randomized tree (ERT), decision tree (DT), random forest (RF), and AdaBoost. Among all models, PseTS-PSSM + RAAA with an extremely randomized tree attained the best outcomes. The proposed predictor (AFP-SPTS) boosted the accuracies of AFPs in the literature by 1.82 and 4.1%.

Prediction of antifreeze proteins using machine learning

Living organisms including fishes, microbes, and animals can live in extremely cold weather. To stay alive in cold environments, these species generate antifreeze proteins (AFPs), also referred to as ice-binding proteins. Moreover, AFPs are extensively utilized in many important fields including medical, agricultural, industrial, and biotechnological. Several predictors were constructed to identify AFPs. However, due to the sequence and structural heterogeneity of AFPs, correct identification is still a challenging task. It is highly desirable to develop a more promising predictor. In this research, a novel computational method, named AFP-LXGB has been proposed for prediction of AFPs more precisely. The information is explored by Dipeptide Composition (DPC), Grouped Amino Acid Composition (GAAC), Position Specific Scoring Matrix-Segmentation-Autocorrelation Transformation (Sg-PSSM-ACT), and Pseudo Position Specific Scoring Matrix Tri-Slicing (PseTS-PSSM). Keeping the benefits of ensemble learning, these feature sets are concatenated into different combinations. The best feature set is selected by Extremely Randomized Tree-Recursive Feature Elimination (ERT-RFE). The models are trained by Light eXtreme Gradient Boosting (LXGB), Random Forest (RF), and Extremely Randomized Tree (ERT). Among classifiers, LXGB has obtained the best prediction results. The novel method (AFP-LXGB) improved the accuracies by 3.70% and 4.09% than the best methods. These results verified that AFP-LXGB can predict AFPs more accurately and can participate in a significant role in medical, agricultural, industrial, and biotechnological fields.