Distantly supervised biomedical relation extraction using piecewise attentive convolutional neural network and reinforcement learning

Abstract

Objective There have been various methods to deal with the erroneous training data in distantly supervised relation extraction (RE), however, their performance is still far from satisfaction. We aimed to deal with the insufficient modeling problem on instance-label correlations for predicting biomedical relations using deep learning and reinforcement learning. Materials and Methods In this study, a new computational model called piecewise attentive convolutional neural network and reinforcement learning (PACNN+RL) was proposed to perform RE on distantly supervised data generated from Unified Medical Language System with MEDLINE abstracts and benchmark datasets. In PACNN+RL, PACNN was introduced to encode semantic information of biomedical text, and the RL method with memory backtracking mechanism was leveraged to alleviate the erroneous data issue. Extensive experiments were conducted on 4 biomedical RE tasks. Results The proposed PACNN+RL model achieved competitive performance on 8 biomedical corpora, outperforming most baseline systems. Specifically, PACNN+RL outperformed all baseline methods with the F1-score of 0.5592 on the may-prevent dataset, 0.6666 on the may-treat dataset, and 0.3838 on the DDI corpus, 2011. For the protein-protein interaction RE task, we obtained new state-of-the-art performance on 4 out of 5 benchmark datasets. Conclusions The performance on many distantly supervised biomedical RE tasks was substantially improved, primarily owing to the denoising effect of the proposed model. It is anticipated that PACNN+RL will become a useful tool for large-scale RE and other downstream tasks to facilitate biomedical knowledge acquisition. We also made the demonstration program and source code publicly available at http://112.74.48.115:9000/.

Deciphering the Diversity of Mental Models in Neurodevelopmental Disorders: Knowledge Graph Representation of Public Data Using Natural Language Processing

“…Utilizing this information could provide more meaningful and direct relations between the concepts of different semantic types. We aim to apply the distantly supervised relation extraction approach on each document corpus, which utilizes the UMLS semantic network to obtain diverse relations between different concepts [ 67 , 68 ]. The output of this approach can also be used as training data for deep learning algorithms to train relation extraction models, which would allow us to create KG by processing text corpus not only for the NDD domain but also for any other condition.…”

Recent Advances in Deep Learning for Protein-Protein Interaction Analysis: A Comprehensive Review

“…Lastly, several studies have been dedicated to the application of deep learning in biomedical relation extraction. Notably, Zhu et al [ 147 ] proposed PACNN + RL, a hybrid deep learning and reinforcement learning method for this task. On the other hand, Jovine [ 121 ] used AlphaFold2 and ColabFold to investigate the activation and polymerization of uromodulin, thus showcasing the practical applicability of these methods in biomedicine.…”

Towards Structuring Real-World Data at Scale: Deep Learning for Extracting Key Oncology Information from Clinical Text with Patient-Level Supervision

“…With patient-level supervision from medical registries, our machine learning setting can be regarded as a form of distant supervision or more generally self-supervision [33], as the labels cannot be attributed to a sentence or even a clinical document. However, given the aforementioned complex linguistic phenomena in medical abstraction, we do not generate noisy training examples by associating a label with a specific text span (e.g., individual sentences with the presence of relevant entities), as in standard distant supervision.…”