Prediction of antifreeze proteins using machine learning

Abstract

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.

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

“…According to previous works, evolutionary characteristics have successfully boosted the performance of many biological problems. − Considering this, we formulated the evolutionary patterns via a position-specific scoring matrix (PSSM) employing PSI-BLAST by comparing each sequence with evolutionary sequences in the databank of NCBI’s NR with three iterations and h = 0.001. , Further, slicing and pseudo-notions were extended into PSSM for the extraction of important patterns. The details of these concepts are listed below.…”

“…ERT was widely used for the classification of biological problems such as the identification of N 6 -methyladenosine, 43 anti-tubercular peptides, 44 recognition of antifreeze proteins, 28 classification of DNA-binding proteins, 45 protein hot spot identification, 46 cell-penetrating peptides, 47 and anti-inflammatory peptides. 48 ERT was adopted in medical fields like segmentation of brain tumors 49 and identifying genetic issues.…”

Recent Advances in Machine Learning-Based Models for Prediction of Antiviral Peptides

“…SVM implements grid search approach to find the best values for C and γ parameters to improve the model prediction. Due to promising performance of SVM, it was used for solving many research challenging tasks like protein remote homology detection [ 55 ], protein structure prediction [ 56 ], identification of antifreeze protein [ 57 ], identification of DNA-binding proteins [ 14 ], protein fold recognition, and prediction of promoter [ 58 ]. SVM was applied by Thakur et al [ 1 ], AVCpred [ 4 ], and FIRM-AVP [ 6 ] for prediction of antiviral peptides.…”

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

“…Recently, some machine learning approaches have been presented for the prediction of AFPs from their sequences using different feature descriptors. Examples of a few such methods are AFP-Pred, AFP-PSSM, AFP-pseAAC, i-AFP, RAFP-Pred, AFP-Ensemble, CryoProtect, AFP-LSE, AFP-LXBG, and AFP-SPTS . One of the early machine learning models to classify AFPs was proposed by Kandaswamy et al (AFP-Pred), they used a total of 119 features, which included frequency of hydrophobic, hydrophilic, neutral, polar, and nonpolar amino acids, frequency of the different amino acids in coil, helix, and strand region, and several other physiochemical properties to train the random forest (RF) classifier.…”

“…Recently, some machine learning approaches have been presented for the prediction of AFPs from their sequences using different feature descriptors. Examples of a few such methods are AFP-Pred, 38 AFP-PSSM, 39 AFP-pseAAC, 40 i-AFP, 41 RAFP-Pred, 42 AFP-Ensemble, 43 CryoProtect, 44 AFP-LSE, 45 AFP-LXBG, 46 and AFP-SPTS. 47 Although every method increases the accuracy of AFP prediction, there remains a continuous need for further enhancement to achieve a high accuracy of AFP prediction.…”