Lucila Ohno-Machado

Natural language processing: an introduction

OBJECTIVES To provide an overview and tutorial of natural language processing (NLP) and modern NLP-system design. TARGET AUDIENCE This tutorial targets the medical informatics generalist who has limited acquaintance with the principles behind NLP and/or limited knowledge of the current state of the art. SCOPE We describe the historical evolution of NLP, and summarize common NLP sub-problems in this extensive field. We then provide a synopsis of selected highlights of medical NLP efforts. After providing a brief description of common machine-learning approaches that are being used for diverse NLP sub-problems, we discuss how modern NLP architectures are designed, with a summary of the Apache Foundations Unstructured Information Management Architecture. We finally consider possible future directions for NLP, and reflect on the possible impact of IBM Watson on the medical field.

The use of receiver operating characteristic curves in biomedical informatics

Receiver operating characteristic (ROC) curves are frequently used in biomedical informatics research to evaluate classification and prediction models for decision support, diagnosis, and prognosis. ROC analysis investigates the accuracy of a models ability to separate positive from negative cases (such as predicting the presence or absence of disease), and the results are independent of the prevalence of positive cases in the study population. It is especially useful in evaluating predictive models or other tests that produce output values over a continuous range, since it captures the trade-off between sensitivity and specificity over that range. There are many ways to conduct an ROC analysis. The best approach depends on the experiment; an inappropriate approach can easily lead to incorrect conclusions. In this article, we review the basic concepts of ROC analysis, illustrate their use with sample calculations, make recommendations drawn from the literature, and list readily available software.

Analysis of matched mRNA measurements from two different microarray technologies

MOTIVATION [corrected] The existence of several technologies for measuring gene expression makes the question of cross-technology agreement of measurements an important issue. Cross-platform utilization of data from different technologies has the potential to reduce the need to duplicate experiments but requires corresponding measurements to be comparable. METHODS A comparison of mRNA measurements of 2895 sequence-matched genes in 56 cell lines from the standard panel of 60 cancer cell lines from the National Cancer Institute (NCI 60) was carried out by calculating correlation between matched measurements and calculating concordance between cluster from two high-throughput DNA microarray technologies, Stanford type cDNA microarrays and Affymetrix oligonucleotide microarrays. RESULTS In general, corresponding measurements from the two platforms showed poor correlation. Clusters of genes and cell lines were discordant between the two technologies, suggesting that relative intra-technology relationships were not preserved. GC-content, sequence length, average signal intensity, and an estimator of cross-hybridization were found to be associated with the degree of correlation. This suggests gene-specific, or more correctly probe-specific, factors influencing measurements differently in the two platforms, implying a poor prognosis for a broad utilization of gene expression measurements across platforms.

Genomic Analysis of Mouse Retinal Development

The vertebrate retina is comprised of seven major cell types that are generated in overlapping but well-defined intervals. To identify genes that might regulate retinal development, gene expression in the developing retina was profiled at multiple time points using serial analysis of gene expression (SAGE). The expression patterns of 1,051 genes that showed developmentally dynamic expression by SAGE were investigated using in situ hybridization. A molecular atlas of gene expression in the developing and mature retina was thereby constructed, along with a taxonomic classification of developmental gene expression patterns. Genes were identified that label both temporal and spatial subsets of mitotic progenitor cells. For each developing and mature major retinal cell type, genes selectively expressed in that cell type were identified. The gene expression profiles of retinal Müller glia and mitotic progenitor cells were found to be highly similar, suggesting that Müller glia might serve to produce multiple retinal cell types under the right conditions. In addition, multiple transcripts that were evolutionarily conserved that did not appear to encode open reading frames of more than 100 amino acids in length (“noncoding RNAs”) were found to be dynamically and specifically expressed in developing and mature retinal cell types. Finally, many photoreceptor-enriched genes that mapped to chromosomal intervals containing retinal disease genes were identified. These data serve as a starting point for functional investigations of the roles of these genes in retinal development and physiology.

Methodological ReviewLogistic regression and artificial neural network classification models: a methodology review

Logistic regression and artificial neural networks are the models of choice in many medical data classification tasks. In this review, we summarize the differences and similarities of these models from a technical point of view, and compare them with other machine learning algorithms. We provide considerations useful for critically assessing the quality of the models and the results based on these models. Finally, we summarize our findings on how quality criteria for logistic regression and artificial neural network models are met in a sample of papers from the medical literature.

The GuideLine Interchange Format: A Model for Representing Guidelines

OBJECTIVE To allow exchange of clinical practice guidelines among institutions and computer-based applications. DESIGN The GuideLine Interchange Format (GLIF) specification consists of GLIF model and the GLIF syntax. The GLIF model is an object-oriented representation that consists of a set of classes for guideline entities, attributes for those classes, and data types for the attribute values. The GLIF syntax specifies the format of the test file that contains the encoding. METHODS Researchers from the InterMed Collaboratory at Columbia University, Harvard University (Brigham and Womens Hospital and Massachusetts General Hospital), and Stanford University analyzed four existing guideline systems to derive a set of requirements for guideline representation. The GLIF specification is a consensus representation developed through a brainstorming process. Four clinical guidelines were encoded in GLIF to assess its expressivity and to study the variability that occurs when two people from different sites encode the same guideline. RESULTS The encoders reported that GLIF was adequately expressive. A comparison of the encodings revealed substantial variability. CONCLUSION GLIF was sufficient to model the guidelines for the four conditions that were examined. GLIF needs improvement in standard representation of medical concepts, criterion logic, temporal information, and uncertainty.

Twist1-induced invadopodia formation promotes tumor metastasis.

The Twist1 transcription factor is known to promote tumor metastasis and induce Epithelial-Mesenchymal Transition (EMT). Here, we report that Twist1 is capable of promoting the formation of invadopodia, specialized membrane protrusions for extracellular matrix degradation. Twist1 induces PDGFRα expression, which in turn activates Src, to promote invadopodia formation. We show that Twist1 and PDGFRα are central mediators of invadopodia formation in response to various EMT-inducing signals. Induction of PDGFRα and invadopodia is essential for Twist1 to promote tumor metastasis. Consistent with PDGFRα being a direct transcriptional target of Twist1, coexpression of Twist1 and PDGFRα predicts poor survival in breast tumor patients. Therefore, invadopodia-mediated matrix degradation is a key function of Twist1 in promoting tumor metastasis.

Big Data In Health Care: Using Analytics To Identify And Manage High-Risk And High-Cost Patients

The US health care system is rapidly adopting electronic health records, which will dramatically increase the quantity of clinical data that are available electronically. Simultaneously, rapid progress has been made in clinical analytics--techniques for analyzing large quantities of data and gleaning new insights from that analysis--which is part of what is known as big data. As a result, there are unprecedented opportunities to use big data to reduce the costs of health care in the United States. We present six use cases--that is, key examples--where some of the clearest opportunities exist to reduce costs through the use of big data: high-cost patients, readmissions, triage, decompensation (when a patients condition worsens), adverse events, and treatment optimization for diseases affecting multiple organ systems. We discuss the types of insights that are likely to emerge from clinical analytics, the types of data needed to obtain such insights, and the infrastructure--analytics, algorithms, registries, assessment scores, monitoring devices, and so forth--that organizations will need to perform the necessary analyses and to implement changes that will improve care while reducing costs. Our findings have policy implications for regulatory oversight, ways to address privacy concerns, and the support of research on analytics.

Snail2 is an Essential Mediator of Twist1-Induced Epithelial Mesenchymal Transition and Metastasis

To metastasize, carcinoma cells must attenuate cell-cell adhesion to disseminate into distant organs. A group of transcription factors, including Twist1, Snail1, Snail2, ZEB1, and ZEB2, have been shown to induce epithelial mesenchymal transition (EMT), thus promoting tumor dissemination. However, it is unknown whether these transcription factors function independently or coordinately to activate the EMT program. Here we report that direct induction of Snail2 is essential for Twist1 to induce EMT. Snail2 knockdown completely blocks the ability of Twist1 to suppress E-cadherin transcription. Twist1 binds to an evolutionarily conserved E-box on the proximate Snail2 promoter to induce its transcription. Snail2 induction is essential for Twist1-induced cell invasion and distant metastasis in mice. In human breast tumors, the expression of Twist1 and Snail2 is highly correlated. Together, our results show that Twist1 needs to induce Snail2 to suppress the epithelial branch of the EMT program and that Twist1 and Snail2 act together to promote EMT and tumor metastasis.

A Comparison of Machine Learning Methods for the Diagnosis of Pigmented Skin Lesions

We analyze the discriminatory power of k-nearest neighbors, logistic regression, artificial neural networks (ANNs), decision tress, and support vector machines (SVMs) on the task of classifying pigmented skin lesions as common nevi, dysplastic nevi, or melanoma. Three different classification tasks were used as benchmarks: the dichotomous problem of distinguishing common nevi from dysplastic nevi and melanoma, the dichotomous problem of distinguishing melanoma from common and dysplastic nevi, and the trichotomous problem of correctly distinguishing all three classes. Using ROC analysis to measure the discriminatory power of the methods shows that excellent results for specific classification problems in the domain of pigmented skin lesions can be achieved with machine-learning methods. On both dichotomous and trichotomous tasks, logistic regression, ANNs, and SVMs performed on about the same level, with k-nearest neighbors and decision trees performing worse.