As more clinical information with increasing diversity become available for analysis, a large number of features can be constructed and leveraged for predictive modeling. Feature selection is a classic analytic component that faces new challenges due to the new applications: How to handle a diverse set of high dimensional features? How to select features with high predictive power, but low redundant information? How to design methods that can select globally optimal features with theoretical guarantee? How to incorporate and extend existing knowledge driven approach? In this paper, we present Scalable Orthogonal Regression (SOR), an optimization-based feature selection method with the following novelties: 1) Scalability: SOR achieves nearly linear scale-up with respect to the number of input features and the number of samples; 2) Optimality: SOR is formulated as an alternative convex optimization problem with theoretical convergence and global optimality guarantee; 3) Low-redundancy: thanks to the orthogonality objective, SOR is designed specifically to select less redundant features without sacrificing quality; 4) Extendability: SOR can enhance an existing set of preselected features by adding additional features that complement the existing feature set but still with strong predictive power. We present evaluation results showing that SOR consistently outperforms state of the art feature selection methods in a range of quality metrics on several real world data sets. We demonstrate a case study of a large-scale clinical application for predicting early onset of Heart Failure (HF) using real Electronic Health Records (EHRs) data of over 10K patients for over 7 years. Leveraging SOR, we are able to construct accurate and robust predictive models and derive potential clinical insights.

BACKGROUND: Time course microarray profiles examine the expression of genes over a time domain. They are necessary in order to determine the complete set of genes that are dynamically expressed under given conditions, and to determine the interaction between these genes. Because of cost and resource issues, most time series datasets contain less than 9 points and there are few tools available geared towards the analysis of this type of data. RESULTS: To this end, we introduce a platform for Processing Expression of Short Time Series (PESTS). It was designed with a focus on usability and interpretability of analyses for the researcher. As such, it implements several standard techniques for comparability as well as visualization functions. However, it is designed specifically for the unique methods we have developed for significance analysis, multiple test correction and clustering of short time series data. The central tenet of these methods is the use of biologically relevant features for analysis. Features summarize short gene expression profiles, inherently incorporate dependence across time, and allow for both full description of the examined curve and missing data points. CONCLUSIONS: PESTS is fully generalizable to other types of time series analyses. PESTS implements novel methods as well as several standard techniques for comparability and visualization functions. These features and functionality make PESTS a valuable resource for a researcher's toolkit. PESTS is available to download for free to academic and non-profit users at http://www.mailman.columbia.edu/academic-departments/biostatistics/research-service/software-development.

OBJECTIVE: The US Vaccine Adverse Event Reporting System (VAERS) collects spontaneous reports of adverse events following vaccination. Medical officers review the reports and often apply standardized case definitions, such as those developed by the Brighton Collaboration. Our objective was to demonstrate a multi-level text mining approach for automated text classification of VAERS reports that could potentially reduce human workload. DESIGN: We selected 6034 VAERS reports for H1N1 vaccine that were classified by medical officers as potentially positive (N(pos)=237) or negative for anaphylaxis. We created a categorized corpus of text files that included the class label and the symptom text field of each report. A validation set of 1100 labeled text files was also used. Text mining techniques were applied to extract three feature sets for important keywords, low- and high-level patterns. A rule-based classifier processed the high-level feature representation, while several machine learning classifiers were trained for the remaining two feature representations. MEASUREMENTS: Classifiers' performance was evaluated by macro-averaging recall, precision, and F-measure, and Friedman's test; misclassification error rate analysis was also performed. RESULTS: Rule-based classifier, boosted trees, and weighted support vector machines performed well in terms of macro-recall, however at the expense of a higher mean misclassification error rate. The rule-based classifier performed very well in terms of average sensitivity and specificity (79.05% and 94.80%, respectively). CONCLUSION: Our validated results showed the possibility of developing effective medical text classifiers for VAERS reports by combining text mining with informative feature selection; this strategy has the potential to reduce reviewer workload considerably.

Advances in high-throughput and mass-storage technologies have led to an information explosion in both biology and medicine, presenting novel challenges for analysis and modeling. With regards to multivariate analysis techniques such as clustering, classification, and regression, large datasets present unique and often misunderstood challenges. The authors' goal is to provide a discussion of the salient problems encountered in the analysis of large datasets as they relate to modeling and inference to inform a principled and generalizable analysis and highlight the interdisciplinary nature of these challenges. The authors present a detailed study of germane issues including high dimensionality, multiple testing, scientific significance, dependence, information measurement, and information management with a focus on appropriate methodologies available to address these concerns. A firm understanding of the challenges and statistical technology involved ultimately contributes to better science. The authors further suggest that the community consider facilitating discussion through interdisciplinary panels, invited papers and curriculum enhancement to establish guidelines for analysis and reporting.

OBJECTIVE It is vital to detect the full safety profile of a drug throughout its market life. Current pharmacovigilance systems still have substantial limitations, however. The objective of our work is to demonstrate the feasibility of using natural language processing (NLP), the comprehensive Electronic Health Record (EHR), and association statistics for pharmacovigilance purposes. DESIGN Narrative discharge summaries were collected from the Clinical Information System at New York Presbyterian Hospital (NYPH). MedLEE, an NLP system, was applied to the collection to identify medication events and entities which could be potential adverse drug events (ADEs). Co-occurrence statistics with adjusted volume tests were used to detect associations between the two types of entities, to calculate the strengths of the associations, and to determine their cutoff thresholds. Seven drugs/drug classes (ibuprofen, morphine, warfarin, bupropion, paroxetine, rosiglitazone, ACE inhibitors) with known ADEs were selected to evaluate the system. RESULTS One hundred thirty-two potential ADEs were found to be associated with the 7 drugs. Overall recall and precision were 0.75 and 0.31 for known ADEs respectively. Importantly, qualitative evaluation using historic roll back design suggested that novel ADEs could be detected using our system. CONCLUSIONS This study provides a framework for the development of active, high-throughput and prospective systems which could potentially unveil drug safety profiles throughout their entire market life. Our results demonstrate that the framework is feasible although there are some challenging issues. To the best of our knowledge, this is the first study using comprehensive unstructured data from the EHR for pharmacovigilance.

OBJECTIVE: Explore the automated acquisition of knowledge in biomedical and clinical documents using text mining and statistical techniques to identify disease-drug associations. DESIGN: Biomedical literature and clinical narratives from the patient record were mined to gather knowledge about disease-drug associations. Two NLP systems, BioMedLEE and MedLEE, were applied to Medline articles and discharge summaries, respectively. Disease and drug entities were identified using the NLP systems in addition to MeSH annotations for the Medline articles. Focusing on eight diseases, co-occurrence statistics were applied to compute and evaluate the strength of association between each disease and relevant drugs. RESULTS: Ranked lists of disease-drug pairs were generated and cutoffs calculated for identifying stronger associations among these pairs for further analysis. Differences and similarities between the text sources (i.e., biomedical literature and patient record) and annotations (i.e., MeSH and NLP-extracted UMLS concepts) with regards to disease-drug knowledge were observed. CONCLUSION: This paper presents a method for acquiring disease-specific knowledge and a feasibility study of the method. The method is based on applying a combination of NLP and statistical techniques to both biomedical and clinical documents. The approach enabled extraction of knowledge about the drugs clinicians are using for patients with specific diseases based on the patient record, while it is also acquired knowledge of drugs frequently involved in controlled trials for those same diseases. In comparing the disease-drug associations, we found the results to be appropriate: the two text sources contained consistent as well as complementary knowledge, and manual review of the top five disease-drug associations by a medical expert supported their correctness across the diseases.

MOTIVATION: The ambiguity of biomedical entities, particularly of gene symbols, is a big challenge for text-mining systems in the biomedical domain. Existing knowledge sources, such as Entrez Gene and the MEDLINE database, contain information concerning the characteristics of a particular gene that could be used to disambiguate gene symbols. RESULTS: For each gene, we create a profile with different types of information automatically extracted from related MEDLINE abstracts and readily available annotated knowledge sources. We apply the gene profiles to the disambiguation task via an information retrieval method, which ranks the similarity scores between the context where the ambiguous gene is mentioned, and candidate gene profiles. The gene profile with the highest similarity score is then chosen as the correct sense. We evaluated the method on three automatically generated testing sets of mouse, fly and yeast organisms, respectively. The method achieved the highest precision of 93.9% for the mouse, 77.8% for the fly and 89.5% for the yeast. AVAILABILITY: The testing data sets and disambiguation programs are available at http://www.dbmi.columbia.edu/~hux7002/gsd2006

BACKGROUND: Word sense disambiguation (WSD) is critical in the biomedical domain for improving the precision of natural language processing (NLP), text mining, and information retrieval systems because ambiguous words negatively impact accurate access to literature containing biomolecular entities, such as genes, proteins, cells, diseases, and other important entities. Automated techniques have been developed that address the WSD problem for a number of text processing situations, but the problem is still a challenging one. Supervised WSD machine learning (ML) methods have been applied in the biomedical domain and have shown promising results, but the results typically incorporate a number of confounding factors, and it is problematic to truly understand the effectiveness and generalizability of the methods because these factors interact with each other and affect the final results. Thus, there is a need to explicitly address the factors and to systematically quantify their effects on performance. RESULTS: Experiments were designed to measure the effect of "sample size" (i.e. size of the datasets), "sense distribution" (i.e. the distribution of the different meanings of the ambiguous word) and "degree of difficulty" (i.e. the measure of the distances between the meanings of the senses of an ambiguous word) on the performance of WSD classifiers. Support Vector Machine (SVM) classifiers were applied to an automatically generated data set containing four ambiguous biomedical abbreviations: BPD, BSA, PCA, and RSV, which were chosen because of varying degrees of differences in their respective senses. Results showed that: 1) increasing the sample size generally reduced the error rate, but this was limited mainly to well-separated senses (i.e. cases where the distances between the senses were large); in difficult cases an unusually large increase in sample size was needed to increase performance slightly, which was impractical, 2) the sense distribution did not have an effect on performance when the senses were separable, 3) when there was a majority sense of over 90%, the WSD classifier was not better than use of the simple majority sense, 4) error rates were proportional to the similarity of senses, and 5) there was no statistical difference between results when using a 5-fold or 10-fold cross-validation method. Other issues that impact performance are also enumerated. CONCLUSION: Several different independent aspects affect performance when using ML techniques for WSD. We found that combining them into one single result obscures understanding of the underlying methods. Although we studied only four abbreviations, we utilized a well-established statistical method that guarantees the results are likely to be generalizable for abbreviations with similar characteristics. The results of our experiments show that in order to understand the performance of these ML methods it is critical that papers report on the baseline performance, the distribution and sample size of the senses in the datasets, and the standard deviation or confidence intervals. In addition, papers should also characterize the difficulty of the WSD task, the WSD situations addressed and not addressed, as well as the ML methods and features used. This should lead to an improved understanding of the generalizablility and the limitations of the methodology.

This paper brings together methods from two different disciplines: statistics and machine learning. We address the problem of estimating the variance of cross-validation (CV) estimators of the generalization error. In particular, we approach the problem of variance estimation of the CV estimators of generalization error as a problem in approximating the moments of a statistic. The approximation illustrates the role of training and test sets in the performance of the algorithm. It provides a unifying approach to evaluation of various methods used in obtaining training and test sets and it takes into account the variability due to different training and test sets. For the simple problem of predicting the sample mean and in the case of smooth loss functions, we show that the variance of the CV estimator of the generalization error is a function of the moments of the random variables Y=Card(Sj ∩ Sj') and Y*=Card(Sjc ∩ Sj'c), where Sj, Sj' are two training sets, and Sjc, Sj'c are the corresponding test sets. We prove that the distribution of Y and Y* is hypergeometric and we compare our estimator with the one proposed by Nadeau and Bengio (2003). We extend these results in the regression case and the case of absolute error loss, and indicate how the methods can be extended to the classification case. We illustrate the results through simulation.