Korvigo, I. | bioRxiv Science

bioRxiv2018

Putting hands to rest: efficient deep CNN-RNN architecture for chemical named entity recognition with no hand-crafted rules.

Chemical named entity recognition (NER) is an active field of research in biomedical natural language processing. To facilitate the development of new and superior chemical NER systems, BioCreative released the CHEMDNER corpus, an extensive dataset of diverse manually annotated chemical entities. Most of the systems trained on the corpus rely on complicated hand-crafted rules or curated databases for data preprocessing, feature extraction and output post-processing, though modern machine learning algorithms, such as deep neural networks, can automatically design the rules with little to none human intervention. Here we explored this approach by experimenting with various deep learning architectures for targeted tokenisation and named entity recognition. Our final model, based on a combination of convolutional and stateful recurrent neural networks with attention-like loops and hybrid word-and character-level embeddings, reaches near human-level performance on the testing dataset with no manually asserted rules. To make our model easily accessible for standalone use and integration in third-party software, weve developed a Python package with a minimalistic user interface.

bioinformatics↗

bioRxiv2017

Generalising Better: Applying Deep Learning To Integrate Deleteriousness Prediction Scores For Whole-Exome SNV Studies

Many automatic classifiers were introduced to aid inference of phenotypical effects of uncategorised nsSNVs (nonsynonymous Single Nucleotide Variations) in theoretical and medical applications. Lately, several meta-estimators have been proposed that combine different predictors, such as PolyPhen and SIFT, to integrate more information in a single score. Although many advances have been made in feature design and machine learning algorithms used, the shortage of high-quality reference data along with the bias towards intensively studied in vitro models call for improved generalisation ability in order to further increase classification accuracy and handle records with insufficient data. Since a meta-estimator basically combines different scoring systems with highly complicated nonlinear relationships, we investigated how deep learning (supervised and unsupervised), which is particularly efficient at discovering hierarchies of features, can improve classification performance. While it is believed that one should only use deep learning for high-dimensional input spaces and other models (logistic regression, support vector machines, Bayesian classifiers, etc) for simpler inputs, we still believe that the ability of neural networks to discover intricate structure in highly heterogenous datasets can aid a meta-estimator. We compare the performance with various popular predictors, many of which are recommended by the American College of Medical Genetics and Genomics (ACMG), as well as available deep learning-based predictors. Thanks to hardware acceleration we were able to use a computationally expensive genetic algorithm to stochastically optimise hyper-parameters over many generations. Overfitting was hindered by noise injection and dropout, limiting coadaptation of hidden units. Although we stress that this work was not conceived as a tool comparison, but rather an exploration of the possibilities of deep learning application in ensemble scores, our results show that even relatively simple modern neural networks can significantly improve both prediction accuracy and coverage. We provide open-access to our finest model at http://score.generesearch.ru.

bioinformatics↗