Brett K. Beaulieu-Jones

Opportunities and obstacles for deep learning in biology and medicine

Deep learning describes a class of machine learning algorithms that are capable of combining raw inputs into layers of intermediate features. These algorithms have recently shown impressive results across a variety of domains. Biology and medicine are data-rich disciplines, but the data are complex and often ill-understood. Hence, deep learning techniques may be particularly well suited to solve problems of these fields. We examine applications of deep learning to a variety of biomedical problems—patient classification, fundamental biological processes and treatment of patients—and discuss whether deep learning will be able to transform these tasks or if the biomedical sphere poses unique challenges. Following from an extensive literature review, we find that deep learning has yet to revolutionize biomedicine or definitively resolve any of the most pressing challenges in the field, but promising advances have been made on the prior state of the art. Even though improvements over previous baselines have been modest in general, the recent progress indicates that deep learning methods will provide valuable means for speeding up or aiding human investigation. Though progress has been made linking a specific neural networks prediction to input features, understanding how users should interpret these models to make testable hypotheses about the system under study remains an open challenge. Furthermore, the limited amount of labelled data for training presents problems in some domains, as do legal and privacy constraints on work with sensitive health records. Nonetheless, we foresee deep learning enabling changes at both bench and bedside with the potential to transform several areas of biology and medicine.

Reproducibility of computational workflows is automated using continuous analysis

Replication, validation and extension of experiments are crucial for scientific progress. Computational experiments are scriptable and should be easy to reproduce. However, computational analyses are designed and run in a specific computing environment, which may be difficult or impossible to match using written instructions. We report the development of continuous analysis, a workflow that enables reproducible computational analyses. Continuous analysis combines Docker, a container technology akin to virtual machines, with continuous integration, a software development technique, to automatically rerun a computational analysis whenever updates or improvements are made to source code or data. This enables researchers to reproduce results without contacting the study authors. Continuous analysis allows reviewers, editors or readers to verify reproducibility without manually downloading and rerunning code and can provide an audit trail for analyses of data that cannot be shared.

Semi-supervised learning of the electronic health record for phenotype stratification

Patient interactions with health care providers result in entries to electronic health records (EHRs). EHRs were built for clinical and billing purposes but contain many data points about an individual. Mining these records provides opportunities to extract electronic phenotypes, which can be paired with genetic data to identify genes underlying common human diseases. This task remains challenging: high quality phenotyping is costly and requires physician review; many fields in the records are sparsely filled; and our definitions of diseases are continuing to improve over time. Here we develop and evaluate a semi-supervised learning method for EHR phenotype extraction using denoising autoencoders for phenotype stratification. By combining denoising autoencoders with random forests we find classification improvements across multiple simulation models and improved survival prediction in ALS clinical trial data. This is particularly evident in cases where only a small number of patients have high quality phenotypes, a common scenario in EHR-based research. Denoising autoencoders perform dimensionality reduction enabling visualization and clustering for the discovery of new subtypes of disease. This method represents a promising approach to clarify disease subtypes and improve genotype-phenotype association studies that leverage EHRs.

Missing Data Imputation in the Electronic Health Record Using Deeply Learned Autoencoders.

Electronic health records (EHRs) have become a vital source of patient outcome data but the widespread prevalence of missing data presents a major challenge. Different causes of missing data in the EHR data may introduce unintentional bias. Here, we compare the effectiveness of popular multiple imputation strategies with a deeply learned autoencoder using the Pooled Resource Open-Access ALS Clinical Trials Database (PRO-ACT). To evaluate performance, we examined imputation accuracy for known values simulated to be either missing completely at random or missing not at random. We also compared ALS disease progression prediction across different imputation models. Autoencoders showed strong performance for imputation accuracy and contributed to the strongest disease progression predictor. Finally, we show that despite clinical heterogeneity, ALS disease progression appears homogenous with time from onset being the most important predictor.

Privacy-preserving generative deep neural networks support clinical data sharing

Though it is widely recognized that data sharing enables faster scientific progress, the sensible need to protect participant privacy hampers this practice in medicine. We train deep neural networks that generate synthetic subjects closely resembling study participants. Using the SPRINT trial as an example, we show that machine-learning models built from simulated participants generalize to the original dataset. We incorporate differential privacy, which offers strong guarantees on the likelihood that a subject could be identified as a member of the trial. Investigators who have compiled a dataset can use our method to provide a freely accessible public version that enables other scientists to perform discovery-oriented analyses. Generated data can be released alongside analytical code to enable fully reproducible workflows, even when privacy is a concern. By addressing data sharing challenges, deep neural networks can facilitate the rigorous and reproducible investigation of clinical datasets. One Sentence Summary Deep neural networks can generate shareable biomedical data to allow reanalysis while preserving the privacy of study participants.

Mapping Patient Trajectories using Longitudinal Extraction and Deep Learning in the MIMIC-III Critical Care Database

Electronic Health Records (EHRs) contain a wealth of patient data useful to biomedical researchers. At present, both the extraction of data and methods for analyses are frequently designed to work with a single snapshot of a patients record. Health care providers often perform and record actions in small batches over time. By extracting these care events, a sequence can be formed providing a trajectory for a patients interactions with the health care system. These care events also offer a basic heuristic for the level of attention a patient receives from health care providers. We show that is possible to learn meaningful embeddings from these care events using two deep learning techniques, unsupervised autoencoders and long short-term memory networks. We compare these methods to traditional machine learning methods which require a point in time snapshot to be extracted from an EHR.

Semi-Supervised Learning of the Electronic Health Record with Denoising Autoencoders for Phenotype Stratification

Patient interactions with health care providers result in entries to electronic health records (EHRs). EHRs were built for clinical and billing purposes but contain many data points about an individual. Mining these records provides opportunities to extract electronic phenotypes, which can be paired with genetic data to identify genes underlying common human diseases. This task remains challenging: high quality phenotyping is costly and requires physician review; many fields in the records are sparsely filled; and our definitions of diseases are continuing to improve over time. Here we develop and evaluate a semi-supervised learning method for EHR phenotype extraction using denoising autoencoders for phenotype stratification. By combining denoising autoencoders with random forests we find classification improvements across multiple simulation models and improved survival prediction in ALS clinical trial data. This is particularly evident in cases where only a small number of patients have high quality phenotypes, a common scenario in EHR-based research. Denoising autoencoders perform dimensionality reduction enabling visualization and clustering for the discovery of new subtypes of disease. This method represents a promising approach to clarify disease subtypes and improve genotype-phenotype association studies that leverage EHRs. GRAPHICAL ABSTRACT HIGHLIGHTS Denoising autoencoders (DAs) can model electronic health records. Semi-supervised learning with DAs improves ALS patient survival predictions. DAs improve patient cluster visualization through dimensionality reduction.

Machine Learning for Structured Clinical Data

Research is a tertiary priority in the EHR, where the priorities are patient care and billing. Because of this, the data is not standardized or formatted in a manner easily adapted to machine learning approaches. Data may be missing for a large variety of reasons ranging from individual input styles to differences in clinical decision making, for example, which lab tests to issue. Few patients are annotated at a research quality, limiting sample size and presenting a moving gold standard. Patient progression over time is key to understanding many diseases but many machine learning algorithms require a snapshot, at a single time point, to create a usable vector form. Furthermore, algorithms that produce black box results do not provide the interpretability required for clinical adoption. This chapter discusses these challenges and others in applying machine learning techniques to the structured EHR (i.e. Patient Demographics, Family History, Medication Information, Vital Signs, Laboratory Tests, Genetic Testing). It does not cover feature extraction from additional sources such as imaging data or free text patient notes but the approaches discussed can include features extracted from these sources.

Characterizing and Managing Missing Structured Data in Electronic Health Records: Data Analysis

Background Missing data is a challenge for all studies; however, this is especially true for electronic health record (EHR)-based analyses. Failure to appropriately consider missing data can lead to biased results. While there has been extensive theoretical work on imputation, and many sophisticated methods are now available, it remains quite challenging for researchers to implement these methods appropriately. Here, we provide detailed procedures for when and how to conduct imputation of EHR laboratory results. Objective The objective of this study was to demonstrate how the mechanism of missingness can be assessed, evaluate the performance of a variety of imputation methods, and describe some of the most frequent problems that can be encountered. Methods We analyzed clinical laboratory measures from 602,366 patients in the EHR of Geisinger Health System in Pennsylvania, USA. Using these data, we constructed a representative set of complete cases and assessed the performance of 12 different imputation methods for missing data that was simulated based on 4 mechanisms of missingness (missing completely at random, missing not at random, missing at random, and real data modelling). Results Our results showed that several methods, including variations of Multivariate Imputation by Chained Equations (MICE) and softImpute, consistently imputed missing values with low error; however, only a subset of the MICE methods was suitable for multiple imputation. Conclusions The analyses we describe provide an outline of considerations for dealing with missing EHR data, steps that researchers can perform to characterize missingness within their own data, and an evaluation of methods that can be applied to impute clinical data. While the performance of methods may vary between datasets, the process we describe can be generalized to the majority of structured data types that exist in EHRs, and all of our methods and code are publicly available.

Characterizing and Managing Missing Structured Data in Electronic Health Records

Missing data is a challenge for all studies; however, this is especially true for electronic health record (EHR) based analyses. Failure to appropriately consider missing data can lead to biased results. Here, we provide detailed procedures for when and how to conduct imputation of EHR data. We demonstrate how the mechanism of missingness can be assessed, evaluate the performance of a variety of imputation methods, and describe some of the most frequent problems that can be encountered. We analyzed clinical lab measures from 602,366 patients in the Geisinger Health System EHR. Using these data, we constructed a representative set of complete cases and assessed the performance of 12 different imputation methods for missing data that was simulated based on 4 mechanisms of missingness. Our results show that several methods including variations of Multivariate Imputation by Chained Equations (MICE) and softImpute consistently imputed missing values with low error; however, only a subset of the MICE methods were suitable for multiple imputation. The analyses described provide an outline of considerations for dealing with missing EHR data, steps that researchers can perform to characterize missingness within their own data, and an evaluation of methods that can be applied to impute clinical data. While the performance of methods may vary between datasets, the process we describe can be generalized to the majority of structured data types that exist in EHRs and all of our methods and code are publicly available.