Large-Scale Optimization-Based Classification Models in Medicine and Biology

Abstract

Abstract-We present novel optimization-based classification models that are general purpose and suitable for developing predictive rules for large heterogeneous biological and medical data sets. Our predictive model simultaneously incorporates (1) the ability to classify any number of distinct groups; (2) the ability to incorporate heterogeneous types of attributes as input; (3) a high-dimensional data transformation that eliminates noise and errors in biological data; (4) the ability to incorporate constraints to limit the rate of misclassification, and a reserved-judgment region that provides a safeguard against over-training (which tends to lead to high misclassification rates from the resulting predictive rule); and (5) successive multi-stage classification capability to handle data points placed in the reserved-judgment region. To illustrate the power and flexibility of the classification model and solution engine, and its multi-group prediction capability, application of the predictive model to a broad class of biological and medical problems is described. Applications include: the differential diagnosis of the type of erythemato-squamous diseases; predicting presence/absence of heart disease; genomic analysis and prediction of aberrant CpG island meythlation in human cancer; discriminant analysis of motility and morphology data in human lung carcinoma; prediction of ultrasonic cell disruption for drug delivery; identification of tumor shape and volume in treatment of sarcoma; discriminant analysis of biomarkers for prediction of early atherosclerois; fingerprinting of native and angiogenic microvascular networks for early diagnosis of diabetes, aging, macular degeneracy and tumor metastasis; prediction of protein localization sites; and pattern recognition of satellite images in classification of soil types. In all these applications, the predictive model yields correct classification rates ranging from 80 to 100%. This provides motivation for pursuing its use as a medical diagnostic, monitoring and decision-making tool.

Analysis of the consistency of a mixed integer programming-based multi-category constrained discriminant model

“…Since 1996, Lee and her medical colleagues have explored and demonstrated the capability of DAMIP in classifying various types of data arising from real biological and medical problems. DAMIP has been able to consistently maximize the correct classification rate (80%-100% correct rates were obtained) while satisfying pre-set limits on inter-group misclassifications (Gallagher et al 1996(Gallagher et al , 1997Feltus et al 2003Feltus et al , 2006Lee et al 2002Lee et al , 2004Lee 2007aLee , 2007bLee and Wu 2007). In these real applications, beyond reporting the tenfold cross-validation results, the resulting classification rule was also blind tested against new data of unknown group identity and resulted in remarkable rates of correct prediction.…”

“…DAMIP consistently returned superior classification rates compared to other methods (80-100% correct classification rates) while satisfying pre-set limits on inter-group misclassifications (Gallagher et al 1996(Gallagher et al , 1997Feltus et al 2003Feltus et al , 2006Lee et al 2002Lee et al , 2004Lee 2007a, 2007b, Lee and Wu 2007. In these applications, the DAMIP model was solved using a specialized solver, MIPSOL, developed by Lee.…”

“…In addition to heart disease and thyroid disease classification, DAMIP has been applied successfully to prediction of ultrasonic cell disruption for drug delivery (Lee et al 2004), genomic analysis and prediction of aberrant CpG island methylation in human cancer (Feltus et al 2003(Feltus et al , 2006, and identification of tumor shape and volume in treatment of sarcoma (Lee et al 2002) applications. Further, Lee (2007aLee ( , 2007b, Lee and Wu (2007) demonstrated the classification capability of DAMIP over a wide variety of medical and biological problems in which DAMIP offers superior classification results (both in ten-fold cross-validation as well as in blind tests) when compared to other classification approaches.…”

Thymidine kinase and mtDNA depletion in human cardiomyopathy: epigenetic and translational evidence for energy starvation

“…This approach enabled us to identify 131 differentially methylated gene promoters. In stage 2, a binary particle swarm optimization (PSO) algorithm combined with DAMIP was used to identify genes that displayed changes in promoter methylation (11,12,24,34). We reported the results with 100% 10-fold cross validation accuracy for both DCM and NF groups.…”

EM323: a line search based algorithm for solving high-dimensional continuous non-linear optimization problems

“…. ; ng: There has recently been considerable interest in solving instances of (1) of large dimensions (i.e., large values of n), motivated by the emergence of applications in bio-computing, web and data mining (Lee 2007;Nasiri et al 2009). Many current optimization techniques can efficiently solve instances of (1) of small to moderate dimension, involving up to 50 variables.…”