A Classification Model to Predict the Rate of Decline of Kidney Function.

The African American Study of Kidney Disease and Hypertension (AASK), a randomized double-blinded treatment trial, was motivated by the high rate of hypertension-related renal disease in the African-American population and the scarcity of effective therapies. This study describes a pattern-based classification approach to predict the rate of decline of kidney function using surface-enhanced laser desorption ionization/time of flight proteomic data from rapid and slow progressors classified by rate of change in glomerular filtration rate. An accurate classification model consisting of 7 out of 5,751 serum proteomic features is constructed by applying the logical analysis of data (LAD) methodology. On cross-validation by 10-folding, the model was shown to have an accuracy of 80.6 ± 0.11%, sensitivity of 78.4 ± 0.17%, and specificity of 78.5 ± 0.16%. The LAD discriminant is used to identify the patients in different risk groups. The LAD risk scores assigned to 116 AASK patients generated a receiver operating curves curve with AUC 0.899 (CI 0.845-0.953) and outperforms the risk scores assigned by proteinuria, one of the best predictors of chronic kidney disease progression.

Comprehensive peroxidase-based hematologic profiling for the prediction of 1-year myocardial infarction and death.

BACKGROUND: Recognition of biological patterns holds promise for improved identification of patients at risk for myocardial infarction (MI) and death. We hypothesized that identifying high- and low-risk patterns from a broad spectrum of hematologic phenotypic data related to leukocyte peroxidase-, erythrocyte- and platelet-related parameters may better predict future cardiovascular risk in stable cardiac patients than traditional risk factors alone. METHODS AND RESULTS: Stable patients (n=7369) undergoing elective cardiac evaluation at a tertiary care center were enrolled. A model (PEROX) that predicts incident 1-year death and MI was derived from standard clinical data combined with information captured by a high-throughput peroxidase-based hematology analyzer during performance of a complete blood count with differential. The PEROX model was developed using a random sampling of subjects in a derivation cohort (n=5895) and then independently validated in a nonoverlapping validation cohort (n=1474). Twenty-three high-risk (observed in > or =10% of subjects with events) and 24 low-risk (observed in > or =10% of subjects without events) patterns were identified in the derivation cohort. Erythrocyte- and leukocyte (peroxidase)-derived parameters dominated the variables predicting risk of death, whereas variables in MI risk patterns included traditional cardiac risk factors and elements from all blood cell lineages. Within the validation cohort, the PEROX model demonstrated superior prognostic accuracy (78%) for 1-year risk of death or MI compared with traditional risk factors alone (67%). Furthermore, the PEROX model reclassified 23.5% (P<0.001) of patients to different risk categories for death/MI when added to traditional risk factors. CONCLUSIONS: Comprehensive pattern recognition of high- and low-risk clusters of clinical, biochemical, and hematologic parameters provided incremental prognostic value in stable patients having elective diagnostic cardiac catheterization for 1-year risks of death and MI.

Logical Analysis of Data (LAD) model for the early diagnosis of acute ischemic stroke.

BACKGROUND: Strokes are a leading cause of morbidity and the first cause of adult disability in the United States. Currently, no biomarkers are being used clinically to diagnose acute ischemic stroke. A diagnostic test using a blood sample from a patient would potentially be beneficial in treating the disease. RESULTS: A classification approach is described for differentiating between proteomic samples of stroke patients and controls, and a second novel predictive model is developed for predicting the severity of stroke as measured by the National Institutes of Health Stroke Scale (NIHSS). The models were constructed by applying the Logical Analysis of Data (LAD) methodology to the mass peak profiles of 48 stroke patients and 32 controls. The classification model was shown to have an accuracy of 75% when tested on an independent validation set of 35 stroke patients and 25 controls, while the predictive model exhibited superior performance when compared to alternative algorithms. In spite of their high accuracy, both models are extremely simple and were developed using a common set consisting of only 3 peaks. CONCLUSION: We have successfully identified 3 biomarkers that can detect ischemic stroke with an accuracy of 75%. The performance of the classification model on the validation set and on cross-validation does not deteriorate significantly when compared to that on the training set, indicating the robustness of the model. As in the case of the LAD classification model, the results of the predictive model validate the function constructed on our support-set for approximating the severity scores of stroke patients. The correlation and root mean absolute error of the LAD predictive model are consistently superior to those of the other algorithms used (Support vector machines, C4.5 decision trees, Logistic regression and Multilayer perceptron).

Breast cancer prognosis by combinatorial analysis of gene expression data.

INTRODUCTION: The potential of applying data analysis tools to microarray data for diagnosis and prognosis is illustrated on the recent breast cancer dataset of van 't Veer and coworkers. We re-examine that dataset using the novel technique of logical analysis of data (LAD), with the double objective of discovering patterns characteristic for cases with good or poor outcome, using them for accurate and justifiable predictions; and deriving novel information about the role of genes, the existence of special classes of cases, and other factors. METHOD: Data were analyzed using the combinatorics and optimization-based method of LAD, recently shown to provide highly accurate diagnostic and prognostic systems in cardiology, cancer proteomics, hematology, pulmonology, and other disciplines. RESULTS: LAD identified a subset of 17 of the 25,000 genes, capable of fully distinguishing between patients with poor, respectively good prognoses. An extensive list of 'patterns' or 'combinatorial biomarkers' (that is, combinations of genes and limitations on their expression levels) was generated, and 40 patterns were used to create a prognostic system, shown to have 100% and 92.9% weighted accuracy on the training and test sets, respectively. The prognostic system uses fewer genes than other methods, and has similar or better accuracy than those reported in other studies. Out of the 17 genes identified by LAD, three (respectively, five) were shown to play a significant role in determining poor (respectively, good) prognosis. Two new classes of patients (described by similar sets of covering patterns, gene expression ranges, and clinical features) were discovered. As a by-product of the study, it is shown that the training and the test sets of van 't Veer have differing characteristics. CONCLUSION: The study shows that LAD provides an accurate and fully explanatory prognostic system for breast cancer using genomic data (that is, a system that, in addition to predicting good or poor prognosis, provides an individualized explanation of the reasons for that prognosis for each patient). Moreover, the LAD model provides valuable insights into the roles of individual and combinatorial biomarkers, allows the discovery of new classes of patients, and generates a vast library of biomedical research hypotheses.

A computational approach to predicting cell growth on polymeric biomaterials.

A predictive model that can correlate the chemical composition of a biomaterial with the biological response of cells that are in contact with that biomaterial would represent a major advance and would facilitate the rational design of new biomaterials. As a first step toward this goal, we report here on the use of Logical Analysis of Data (LAD) to model the effect of selected polymer properties on the growth of two different cell types, rat lung fibroblasts (RLF, a transformed cell line), and normal foreskin fibroblasts (NFF, nontransformed human cells), on 112 surfaces obtained from a combinatorially designed library of polymers. LAD is a knowledge extraction methodology, based on using combinatorics, optimization, and Boolean logic. LAD was trained on a subset of 62 polymers and was then used to predict cell growth on 50 previously untested polymers. Experimental validation indicated that LAD correctly predicted the high and low cell growth polymers and found optimal ranges for polymer chemical composition, surface chemistry, and bulk properties. Particularly noteworthy is that LAD correctly identified high-performing polymer surfaces, which surpassed commercial tissue culture polystyrene as growth substratum for normal foreskin fibroblasts. Our results establish the feasibility of using computational modeling of cell growth on flat polymeric surfaces to identify promising "lead" polymers for applications that require either high or low cell growth.

Ovarian cancer detection by logical analysis of proteomic data.

A new type of efficient and accurate proteomic ovarian cancer diagnosis systems is proposed. The system is developed using the combinatorics and optimization-based methodology of logical analysis of data (LAD) to the Ovarian Dataset 8-7-02 (http://clinicalproteomics.steem.com), which updates the one used by Petricoin et al. in The Lancet 2002, 359, 572-577. This mass spectroscopy-generated dataset contains expression profiles of 15 154 peptides defined by their mass/charge ratios (m/z) in serum of 162 ovarian cancer and 91 control cases. Several fully reproducible models using only 7-9 of the 15 154 peptides were constructed, and shown in multiple cross-validation tests (k-folding and leave-one-out) to provide sensitivities and specificities of up to 100%. A special diagnostic system for stage I ovarian cancer patients is shown to have similarly high accuracy. Other results: (i) expressions of peptides with relatively low m/z values in the dataset are shown to be better at distinguishing ovarian cancer cases from controls than those with higher m/z values; (ii) two large groups of patients with a high degree of similarities among their formal (mathematical) profiles are detected; (iii) several peptides with a blocking or promoting effect on ovarian cancer are identified.

Use of the logical analysis of data method for assessing long-term mortality risk after exercise electrocardiography.

BACKGROUND: Logical Analysis of Data is a methodology of mathematical optimization on the basis of the systematic identification of patterns or "syndromes." In this study, we used Logical Analysis of Data for risk stratification and compared it to regression techniques. METHODS AND RESULTS: Using a cohort of 9454 patients referred for exercise testing, Logical Analysis of Data was applied to identify syndromes based on 20 variables. High-risk syndromes were patterns of up to 3 findings associated with >5-fold increase in risk of death, whereas low-risk syndromes were associated with >5-fold decrease. Syndromes were derived on a randomly derived training set of 4722 patients and validated in 4732 others. There were 15 high-risk and 26 low-risk syndromes. A risk score was derived based on the proportion of possible high risk and low risk syndromes present. A value > or =0, meaning the same or a greater proportion of high-risk syndromes, was noted in 979 patients (21%) in the validation set and was predictive of 5-year death (11% versus 1%, hazard ratio 8.3, 95% CI 5.9 to 11.6, P<0.0001), accounting for 67% of events. Calibration of expected versus observed death rates based on Logical Analysis of Data and Cox regression showed that both methods performed very well. CONCLUSION: Using the Logical Analysis of Data method, we identified subsets of patients who had an increased risk and who also accounted for the majority of deaths. Future research is needed to determine how best to use this technique for risk stratification.