Dataset Information

CL-PMI: A Precursor MicroRNA Identification Method Based on Convolutional and Long Short-Term Memory Networks.

ABSTRACT: MicroRNAs (miRNAs) are the major class of gene-regulating molecules that bind mRNAs. They function mainly as translational repressors in mammals. Therefore, how to identify miRNAs is one of the most important problems in medical treatment. Many known pre-miRNAs have a hairpin ring structure containing more structural features, and it is difficult to identify mature miRNAs because of their short length. Therefore, most research focuses on the identification of pre-miRNAs. Most computational models rely on manual feature extraction to identify pre-miRNAs and do not consider the sequential and spatial characteristics of pre-miRNAs, resulting in a loss of information. As the number of unidentified pre-miRNAs is far greater than that of known pre-miRNAs, there is a dataset imbalance problem, which leads to a degradation of the performance of pre-miRNA identification methods. In order to overcome the limitations of existing methods, we propose a pre-miRNA identification algorithm based on a cascaded CNN-LSTM framework, called CL-PMI. We used a convolutional neural network to automatically extract features and obtain pre-miRNA spatial information. We also employed long short-term memory (LSTM) to capture time characteristics of pre-miRNAs and improve attention mechanisms for long-term dependence modeling. Focal loss was used to improve the dataset imbalance. Compared with existing methods, CL-PMI achieved better performance on all datasets. The results demonstrate that this method can effectively identify pre-miRNAs by simultaneously considering their spatial and sequential information, as well as dealing with imbalance in the datasets.

SUBMITTER: Wang H

PROVIDER: S-EPMC6798641 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Experimental techniques for measuring chromatin accessibility are expensive and time consuming, appealing for the development of computational approaches to predict open chromatin regions from DNA sequences. Along this direction, existing methods fall into two classes: one based on handcrafted k -mer features and the other based on convolutional neural networks. Although both categories have shown good performance in specific applications thus far, there still lacks a comprehensive framework to integrate useful k -mer co-occurrence information with recent advances in deep learning.We fill this gap by addressing the problem of chromatin accessibility prediction with a convolutional Long Short-Term Memory (LSTM) network with k -mer embedding. We first split DNA sequences into k -mers and pre-train k -mer embedding vectors based on the co-occurrence matrix of k -mers by using an unsupervised representation learning approach. We then construct a supervised deep learning architecture comprised of an embedding layer, three convolutional layers and a Bidirectional LSTM (BLSTM) layer for feature learning and classification. We demonstrate that our method gains high-quality fixed-length features from variable-length sequences and consistently outperforms baseline methods. We show that k -mer embedding can effectively enhance model performance by exploring different embedding strategies. We also prove the efficacy of both the convolution and the BLSTM layers by comparing two variations of the network architecture. We confirm the robustness of our model to hyper-parameters by performing sensitivity analysis. We hope our method can eventually reinforce our understanding of employing deep learning in genomic studies and shed light on research regarding mechanisms of chromatin accessibility.The source code can be downloaded from https://github.com/minxueric/ismb2017_lstm .tingchen@tsinghua.edu.cn or ruijiang@tsinghua.edu.cn.Supplementary materials are available at Bioinformatics online.

Project description:Network functioning during cognitive tasks is of major interest in Alzheimer's disease (AD). Cognitive functioning in AD includes variable performance in short-term memory (STM). In most studies, the verbal STM functioning in AD patients has been interpreted within the phonological loop subsystem of Baddeley's working memory model. An alternative account considers that domain-general attentional processes explain the involvement of frontoparietal networks in verbal STM beside the functioning of modality-specific subsystems. In this study, we assessed the functional integrity of the dorsal attention network (involved in task-related attention) and the ventral attention network (involved in stimulus-driven attention) by varying attentional control demands in a STM task. Thirty-five AD patients and twenty controls in the seventies performed an fMRI STM task. Variation in load (five versus two items) allowed the dorsal (DAN) and ventral attention networks (VAN) to be studied. ANOVA revealed that performance decreased with increased load in both groups. AD patients performed slightly worse than controls, but accuracy remained above 70% in all patients. Statistical analysis of fMRI brain images revealed DAN activation for high load in both groups. There was no between-group difference or common activation for low compared to high load conditions. Psychophysiological interaction showed a negative relationship between the DAN and the VAN for high versus low load conditions in patients. In conclusion, the DAN remained activated and connected to the VAN in mild AD patients who succeeded in performing an fMRI verbal STM task. DAN was necessary for the task, but not sufficient to reach normal performance. Slightly lower performance in early AD patients compared to controls might be related to maintained bottom-up attention to distractors, to decrease in executive functions, to impaired phonological processing or to reduced capacity in serial order processing.

Project description:BACKGROUND:Recent studies have proposed deep learning techniques, namely recurrent neural networks, to improve biomedical text mining tasks. However, these techniques rarely take advantage of existing domain-specific resources, such as ontologies. In Life and Health Sciences there is a vast and valuable set of such resources publicly available, which are continuously being updated. Biomedical ontologies are nowadays a mainstream approach to formalize existing knowledge about entities, such as genes, chemicals, phenotypes, and disorders. These resources contain supplementary information that may not be yet encoded in training data, particularly in domains with limited labeled data. RESULTS:We propose a new model to detect and classify relations in text, BO-LSTM, that takes advantage of domain-specific ontologies, by representing each entity as the sequence of its ancestors in the ontology. We implemented BO-LSTM as a recurrent neural network with long short-term memory units and using open biomedical ontologies, specifically Chemical Entities of Biological Interest (ChEBI), Human Phenotype, and Gene Ontology. We assessed the performance of BO-LSTM with drug-drug interactions mentioned in a publicly available corpus from an international challenge, composed of 792 drug descriptions and 233 scientific abstracts. By using the domain-specific ontology in addition to word embeddings and WordNet, BO-LSTM improved the F1-score of both the detection and classification of drug-drug interactions, particularly in a document set with a limited number of annotations. We adapted an existing DDI extraction model with our ontology-based method, obtaining a higher F1 score than the original model. Furthermore, we developed and made available a corpus of 228 abstracts annotated with relations between genes and phenotypes, and demonstrated how BO-LSTM can be applied to other types of relations. CONCLUSIONS:Our findings demonstrate that besides the high performance of current deep learning techniques, domain-specific ontologies can still be useful to mitigate the lack of labeled data.

Dataset Information

CL-PMI: A Precursor MicroRNA Identification Method Based on Convolutional and Long Short-Term Memory Networks.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets