A hierarchical multi-label classification method based on neural networks for gene function prediction

Abstract

Gene function prediction is used to assign biological or biochemical functions to genes, which continues to be a challenging problem in modern biology. Genes may exhibit many functions simultaneously, and these functions are organized into a hierarchy, such as a directed acyclic graph (DAG) for Gene Ontology (GO). Because of these characteristics, gene function prediction can be seen as a typical hierarchical multi-label classification (HMC) task. A novel HMC method based on neural networks is proposed in this article for predicting gene function based on GO. The proposed method belongs to a local approach by transferring the HMC task to a set of subtasks. There are three strategies implemented in this method to improve its performance. First, to tackle the imbalanced data set problem when building the training data set for each class, negative instances selecting policy and SMOTE approach are used to preprocess each imbalanced training data set. Second, a particular multi-layer perceptron (MLP) is designed for each node in GO. Third, a post processing method based on the Bayesian network is used to guarantee that the results are consistent with the hierarchy constraint. The experimental results indicate that the proposed HMC-MLPN method is a promising method for gene function prediction based on a comparison with two other state-of-the-art methods.

Machine learning for discovering missing or wrong protein function annotations

“…According to Silla and Freitas [33], the local approach is further divided into three strategies: Local Classifier per Level [3,5,14,25,30], Local Classifier per Node [7,9] and Local Classifier per Parent Node [11,16]. As their name suggest, these strategies train a predictive model for each level, node or parent node of the hierarchy, respectively.…”

“…The work of Feng et al [9] proposed to use the Local Classifier per Node strategy by training one Support Vector Machine for each node of the hierarchy combined with the SMOTE oversampling technique. This work was slightly improved in Feng et al [7] where the Support Vector Machines were replaced by Multi-Layer Perceptron and a post-prediction method based on Bayesian networks was used. Also using Support Vector Machines, the studies of Bi and Kwok [12,20] proposed new loss functions specific for HMC which were optimized using Bayes optimization techniques.…”

Semantic similarity and machine learning with ontologies

“…Further research added several improvements to the DeepGO classifier, both with respect to computational complexity [ 98 ] and by reformulating the ‘hard’ constraints implemented in DeepGO as “soft” constraints using Bayesian networks [ 99 ]. There are several further methods that incorporate hierarchical constraints in artificial neural networks [ 100–102 ], mostly using a variation of the methods employed by ontology-based predictors.…”

A co‐training‐based approach for the hierarchical multi‐label classification of research papers

“…Then, at the test phase, a top-down strategy that surfs the label hierarchy from root to leaf nodes is adopted for deducing the final assigned labels. (Feng, Fu, & Zheng, 2018) propose an LCN-based HMC method that firstly applies a negative instances selecting policy and an oversampling technique for reducing the label imbalance problem in initial training datasets. Then, a particular MLP classifier is learnt for each label in the hierarchy.…”