Biocomputing 2019 - Proceedings Of The Pacific Symposium.

• Author: Altman, Russ B.
• TeilnehmendeR: Dunker, A Keith.; Hunter, Lawrence.; Ritchie, Marylyn D.; Murray, Tiffany A.; Klein, Teri E.
• Place / Publishing House: Singapore : : World Scientific Publishing Company,, 2018.; ©2019.
• Year of Publication: 2018
• Edition: 1st ed.
• Language: English
• Physical Description: 1 online resource (471 pages)

Table of Contents

• Intro
• Preface
• PATTERN RECOGNITION IN BIOMEDICAL DATA: CHALLENGES IN PUTTING BIG DATA TO WORK
• Session introduction
• Introduction
• References
• Learning Contextual Hierarchical Structure of Medical Concepts with Poincairé Embeddings to Clarify Phenotypes
• 1. Introduction
• 2. Methods
• 2.1. Source Code
• 2.2. Data Source
• 2.3. Data Selection and Preprocessing
• 2.3.1. Reference ICD9 Example
• 2.3.2. Real Member Analyses
• 2.4. Poincaré Embeddings
• 2.5. Processing and Evaluating Embeddings
• 3. Results
• 3.1. ICD9 Hierarchy Evaluation
• 3.2. Poincaré Embeddings on 10 Million Members
• 3.3. Comparison with Euclidean Embeddings
• 3.4. Cohort Specific Embeddings
• 4. Discussion and Conclusion
• 5. Acknowledgments
• References
• The Effectiveness of Multitask Learning for Phenotyping with Electronic Health Records Data
• 1. Introduction
• 2. Background
• 2.1. Multitask nets
• 3. Methods
• 3.1. Dataset Construction and Design
• 3.2. Experimental Design
• 4. Experiments and Results
• 4.1. When Does Multitask Learning Improve Performance?
• 4.2. Relationship Between Performance and Number of Tasks
• 4.3. Comparison with Logistic Regression Baseline
• 4.4. Interaction between Phenotype Prevalence and Complexity
• 5. Limitations
• 6. Conclusion
• Acknowledgments
• References
• ODAL: A one-shot distributed algorithm to perform logistic regressions on electronic health records data from multiple clinical sites
• 1. Introduction
• 1.1. Integrate evidence from multiple clinical sites
• 1.2. Distributed Computing
• 2. Material and Method
• 2.1. Clinical Cohort and Motivating Problem
• 2.2. Algorithm
• 2.3. Simulation Design
• 3. Results
• 3.1. Simulation Results
• 3.2. Fetal Loss Prediction via ODAL
• 4. Discussion
• References.
• PVC Detection Using a Convolutional Autoencoder and Random Forest Classifier
• 1. Introduction
• 2. Methods
• 2.1. Data Set and Implementation
• 2.2. Proposed PVC Detection Method
• 2.2.1. Feature Extraction
• 2.2.2. Classification
• 3. Results
• 3.1. Full Database Evaluation
• 3.2. Timing Disturbance Evaluation
• 3.3. Cross-Patient Training Evaluation
• 3.4. Estimated Parameters and Convergence
• 4. Discussion
• References
• Removing Confounding Factors Associated Weights in Deep Neural Networks Improves the Prediction Accuracy for Healthcare Applications
• 1. Introduction
• 2. Related Work
• 3. Confounder Filtering (CF) Method
• 3.1. Overview
• 3.2. Method
• 3.3. Availability
• 4. Experiments
• 4.1. lung adenocarcinoma prediction
• 4.1.1. Data
• 4.1.2. Results
• 4.2. Segmentation on right ventricle(RV) of Heart
• 4.2.1. Data
• 4.2.2. Results
• 4.3. Students' confusion status prediction
• 4.3.1. Data
• 4.3.2. Results
• 4.4. Brain tumor prediction
• 4.4.1. Data
• 4.4.2. Results
• 4.5. Analyses of the method behaviors
• 5. Conclusion
• 6. Acknowledgement
• References
• DeepDom: Predicting protein domain boundary from sequence alone using stacked bidirectional LSTM
• 1. Introduction
• 2. METHODS
• 2.1 Data Set Preparation
• 2.2 Input Encoding
• 2.3 Model Architecture
• 2.4 Evaluation criteria
• 3. RESULTS AND DISCUSSION
• 3.1 Parameter configuration experiments on test data
• 3.2 Comparison with Other Domain Boundary Predictors
• 3.2.1 Free modeling targets from CASP 9
• 3.2.2 Multi-domain targets from CASP 9
• 3.2.3 Discontinuous domain target from CASP 8
• 4. CONCLUSION
• 5. ACKNOWLEDGEMENTS
• REFERENCES
• Res2s2aM: Deep residual network-based model for identifying functional noncoding SNPs in trait-associated regions
• 1. Introduction
• 2. Background theory.
• 3. Dataset for training and testing
• 3.1. Source databases
• 3.2. Dataset generation
• 4. Methods
• 4.1. ResNet architecture in our model
• 4.2. Tandem inputs of forward- and reverse-strand sequences
• 4.3. Biallelic high-level network structure
• 4.4. Incorporating HaploReg SNP annotation features
• 4.5. Training of models
• 5. Results
• 6. Conclusions and discussion
• Acknowledgements
• References
• DNA Steganalysis Using Deep Recurrent Neural Networks
• 1. Introduction
• 2. Background
• 2.1. Notations
• 2.2. Hiding Messages
• 2.3. Determination of Message-Hiding Regions
• 3. Methods
• 3.1. Proposed DNA Steganalysis Principle
• 3.2. Proposed Steganalysis RNN Model
• 4. Results
• 4.1. Dataset
• 4.2. Input Representation
• 4.3. Model Training
• 4.4. Evaluation Procedure
• 4.5. Performance Comparison
• 5. Discussion
• Acknowledgments
• References
• Bi-directional Recurrent Neural Network Models for Geographic Location Extraction in Biomedical Literature
• 1. Introduction
• 2. Related Work
• 3. Methods
• 3.1. Toponym Detection
• 3.1.1. Recurrent Neural Networks
• 3.1.2. LSTM
• 3.1.3. Other Gated RNN Architectures
• 3.1.4. Hyperparameter search and optimization
• 3.2. Toponym Disambiguation
• 3.2.1. Building Geonames Index
• 3.2.2. Searching Geonames Index
• 4. Results and Discussion
• 4.1. Toponym Disambiguation
• 4.2. Toponym Resolution
• 5. Limitations and Future Work
• 6. Conclusion
• Acknowledgments
• Funding
• References
• Automatic Human-like Mining and Constructing Reliable Genetic Association Database with Deep Reinforcement Learning
• 1. Introduction
• 2. Related Work
• 3. Method
• 3.1. Model Framework
• 3.2. Deep Reinforcement Learning for Organizing Actions
• 3.3. Preprocessing and Name Entity Recognition with UMLS
• 3.4. Bidirectional LSTM for Relation Classification.
• 3.5. Algorithm
• 3.6. Implementation Specification
• 4. Experiments
• 4.1. Data
• 4.2. Evaluation
• 4.3. Results
• 4.3.1. Improved Reliability
• 4.3.2. Robustness in Real-world Situations
• 4.3.3. Number of Articles Read
• 5. Conclusions and Future Work
• 6. Acknowledgement
• References
• Estimating classification accuracy in positive-unlabeled learning: characterization and correction strategies
• 1. Introduction
• 2. Methods
• 2.1. Performance measures: definitions and estimation
• 2.2. Positive-unlabeled setting
• 2.3. Performance measure correction
• 3. Experiments and Results
• 3.1. A case study
• 3.2. Data sets
• 3.3. Experimental protocols
• 3.4. Results
• 4. Conclusions
• Acknowledgements
• References
• PLATYPUS: A Multiple-View Learning Predictive Framework for Cancer Drug Sensitivity Prediction
• 1. Introduction
• 2. System and methods
• 2.1. Data
• 2.2. Single views and co-training
• 2.3. Maximizing agreement across views through label assignment
• 3. Results
• 3.1. Preliminary experiments to optimize PLATYPUS performance
• 3.2. Predicting drug sensitivity in cell lines
• 3.3. Key features from PLATYPUS models
• 4. Conclusions
• Acknowledgments
• References
• Computational KIR copy number discovery reveals interaction between inhibitory receptor burden and survival
• 1. Introduction
• 2. Materials and Methods
• 2.1 Data collection
• 2.2 K-mer selection
• 2.3 NGS pipeline and k-mer extraction
• 2.4 Data cleaning
• 2.5 Normalization of k-mer frequencies
• 2.6 Copy number segregation and cutoff selection
• 2.7 Validation of copy number
• 2.8 Survival analysis
• 2.9 Additional immune analysis
• 3. Results and Discussions
• 3.1 Establishing unique k-mers
• 3.2 Varying coverage of KIR region by exome capture kit
• 3.3 Inference of KIR copy number
• 3.4 Population variation of the KIR region.
• 3.5 KIR inhibitory gene burden correlates with survival in cervical and uterine cancer
• 5. Conclusions
• 6. Acknowledgements
• 7. Supplementary Material
• References
• Exploring microRNA Regulation of Cancer with Context-Aware Deep Cancer Classifier
• 1. Introduction
• 2. Data
• 2.1. Preprocessing
• 3. Deep Cancer Classifier
• 3.1. Training &amp
• testing
• 3.2. Parameter tuning
• 3.3. Feature importance
• 4. Results and Discussion
• 4.1. Model selection
• 4.2. Classifier performance
• 4.3. Comparison with other methods
• 4.4. Feature importance
• 5. Conclusion
• References
• Implementing and Evaluating A Gaussian Mixture Framework for Identifying Gene Function from TnSeq Data
• 1. Introduction
• 1.1. TnSeq Motivation and Background
• 1.2. Motivation and New Methods
• 2. Methods
• 2.1. TnSeq Experimental Data
• 2.2. Mixture framework
• 2.3. Classification methods
• 2.3.1. Novel method - EM
• 2.3.2. Current method - t-statistic
• 2.3.3. Bayesian hierarchical model
• 2.3.4. Data partitioning for the Bayesian model
• 2.4. Simulation
• 2.5. Real data
• 3. Results
• 3.1.1. Classification rate
• 3.1.2. False positive rate
• 3.1.3. Positive classification rate
• 3.1.4. Cross entropy
• 3.2. Simulation Results
• 3.3. Comparisons on real data
• 3.4. Software
• 4. Discussion
• References
• SNPs2ChIP: Latent Factors of ChIP-seq to infer functions of non-coding SNPs
• 1. Introduction
• 2. Results
• 2.1. SNPs2ChIP analysis framework overview
• 2.2. Batch normalization of heterogeneous epigenetic features
• 2.3. Latent factor discovery and their biological characterization
• 2.4. SNPs2ChIP identifies relevant functions of the non-coding genome
• 2.4.1. Genome-wide SNPs coverage of the reference datasets
• 2.4.2. Non-coding GWAS SNPs of systemic lupus erythematosus
• 2.4.3. ChIP-seq peaks for vitamin D receptors.
• 2.5. Robustness Analysis in the latent factor identification.