SCORPION is a stacking-based ensemble learning framework for accurate prediction of phage virion proteins

Abstract

Fast and accurate identification of phage virion proteins (PVPs) would greatly aid facilitation of antibacterial drug discovery and development. Although, several research efforts based on machine learning (ML) methods have been made for in silico identification of PVPs, these methods have certain limitations. Therefore, in this study, we propose a new computational approach, termed SCORPION, (StaCking-based Predictior fOR Phage VIrion PrOteiNs), to accurately identify PVPs using only protein primary sequences. Specifically, we explored comprehensive 13 different feature descriptors from different aspects (i.e., compositional information, composition-transition-distribution information, position-specific information and physicochemical properties) with 10 popular ML algorithms to construct a pool of optimal baseline models. These optimal baseline models were then used to generate probabilistic features (PFs) and considered as a new feature vector. Finally, we utilized a two-step feature selection strategy to determine the optimal PF feature vector and used this feature vector to develop a stacked model (SCORPION). Both tenfold cross-validation and independent test results indicate that SCORPION achieves superior predictive performance than its constitute baseline models and existing methods. We anticipate SCORPION will serve as a useful tool for the cost-effective and large-scale screening of new PVPs. The source codes and datasets for this work are available for downloading in the GitHub repository (https://github.com/saeed344/SCORPION).

Potential and limitations of machine meta-learning (ensemble) methods for predicting COVID-19 mortality in a large inhospital Brazilian dataset

“…In a sense, some traditional models, such as regression models, also have a good explainability, as we can assess the coefficients of each attribute to measure how important a feature is. These models however do not measure up in terms of effectiveness when compared to modern tree-based algorithms in many scenarios, especially in cases with larger datasets 33 . Another key difference between these models is that, in the case of regression models, we have to explicitly remove collinear variables, but these variables, even though they might not improve classification performance, still yield valid model explanations.…”

“…The combination of models based on different classification premises potentially made Stacking more robust. If a single classifier makes a wrong prediction, the others can still make corrections, increasing the robustness of the final stacking model [32][33][34][35] .…”

Recent development of machine learning-based methods for the prediction of defensin family and subfamily

“…In order to enhance the performance of SVM classifiers, a grid search strategy was utilized to optimize the two important aspects of the RBF kernel, including C (controls the trade-off between the misclassification rate and margin) and γ (the kernel width parameter). Although SVM often yields satisfactory prediction performances, this method is known as a black-box computation method (Ahmad et al, 2022[ 1 ]; Charoenkwan et al, 2021[ 12 ]; Li et al, 2021[ 33 ]; Wei et al, 2021[ 54 ]).…”

What the protein!? Computational methods for predicting microbial protein functions

“…A common solution is only to store k-mers that appear two or more times and to use a probabilistic data structure like a Bloom Filter or Counting Quotient Filter to remove singleton k-mers [53]. Other valuable numerical features include physicochemical properties, such as isoelectric point, aromaticity, molar extinction coefficient, instability index, molecular weight, polarity and hydrophobicity [54][55][56]. Each of these features can be calculated from the sequence alone using readily available bioinformatics tools [57].…”

“…Once a set of features has been curated, machine learning techniques are applied to differentiate between protein functions. These techniques include Support Vector Machines [58][59][60] and Artificial Neural Networks [54], as well as ensembles of multiple machine learning methods [55,56]. Several function predictors have been released as bioinformatic tools, including PhANNs (Phage Artificial Neural Networks) [54], which can classify phage proteins into one of ten structural classes with 86.2% accuracy.…”