A Shortest Dependency Path Based Convolutional Neural Network for Protein-Protein Relation Extraction

Abstract

The state-of-the-art methods for protein-protein interaction (PPI) extraction are primarily based on kernel methods, and their performances strongly depend on the handcraft features. In this paper, we tackle PPI extraction by using convolutional neural networks (CNN) and propose a shortest dependency path based CNN (sdpCNN) model. The proposed method (1) only takes the sdp and word embedding as input and (2) could avoid bias from feature selection by using CNN. We performed experiments on standard Aimed and BioInfer datasets, and the experimental results demonstrated that our approach outperformed state-of-the-art kernel based methods. In particular, by tracking the sdpCNN model, we find that sdpCNN could extract key features automatically and it is verified that pretrained word embedding is crucial in PPI task.

Named Entity Recognition and Relation Detection for Biomedical Information Extraction

“…The vectors are looked up from pre-trained word and positional vector space on either a single corpus or multiple corpora (Quan et al, 2016 ). Significantly, the majority of deep learning methods use sentence dependency graphs mentioned in the rule-based approach ( Figure 8 ) to extract the shortest path between entities and relations as features for training (Hua and Quan, 2016a , b ; Zhang et al, 2018c ; Li et al, 2019 ). Other studies have used POS tagging, and chunk tagging features in combination with position and dependency paths to improve performance (Peng and Lu, 2017 ).…”

A self-attention based deep learning method for lesion attribute detection from CT reports

“…Owing to the rapid growth of available EHR, deep learning methods for text mining become more appealing recently because of its competitive performance versus traditional methods and its ability to relieve the feature sparsity and engineering issue [7]. For example, both multi-channel dependency-based CNNs [8] and shortest path-based CNNs [9] are well suited for sentence-based relation extraction. It is also generally faster to train a CNN model than other deep learning networks.…”

ProDCoNN: Protein design using a convolutional neural network

“…In deep learning, more efforts are made in selecting the type of DNNs and designing specific architecture for the selected DNN framework. In the past few years, deep learning has seen its applications in computational studies of protein structures, such as protein secondary structure prediction, protein contact map prediction, and protein–protein interaction prediction …”

“…DNNs have also been used for protein design. In 2014, Zhou and co‐workers developed SPIN (sequence profiles by integrated neural network) based on fragment‐derived sequence profiles and structure‐derived energy profiles . Both local and nonlocal features were designed and served as input for a two hidden layer neural network, which contained 51 hidden neurons and one bias.…”