Discipline: Social, Behavioral, and Economic Sciences
Subcategory: Biomedical Engineering
Yazeli E. Cruz-Rivera - University of Puerto Rico at Mayaguez
Co-Author(s): Yaritza M. Santiago, Valerie M. Gonzalez, and Mauricio Cabrera-Rios,The Applied Optimization Group/ Department of Industrial Engineering Clara E. Isaza,The Applied Optimization Group/Ponce Health Sciences University
Biomarkers for Alzheimer’s disease can be used in early detection, diagnosis, prognosis and recurrence. A good biomarking candidate is characterized by its distinct behavior in different states (Pérez-Morales 2014). Genetic relative expression measured using microarrays can be used to detect potential biomarkers. The need for microarray analysis methods that do not depend on user-defined parameters or significance thresholds was previously identified by our research group, thus converging in an optimization driven approach for gene filtering and the determination of potential signaling paths. This analysis pipeline is applied here to the study of Alzheimer’s Disease.
The analysis was carried out using a microarray database by finding coexpression behavior among different genes and measuring it as linear statistical correlation. The potential biomarking genes were found through multiple criteria optimization. From this list, if each gene is represented through a node in a graph, then the undirected arc joining a pair of genes can hold the absolute value of their correlation coefficient. This leads naturally to the Traveling Salesman Problem (TSP) formulation, where the idea is to find the most correlated complete tour. Indeed, this most correlated tour is found through combinatorial optimization and arrives to the optimal tour –the novelty of our work-. Fourteen genes were identified with multiple criteria optimization from the filtering stage. Five of these genes (ND2, SH3KBP1, GFAP, COX3 and RPS10) have been reported with a direct relation to Alzheimer’s or to other neurodegenerative diseases. The rest of the genes (including NUCKS1, PTPRO, DCAF6, MEG3, RPL41 and FTL) have been associated to processes recently studied in the context of Alzheimer’s disease. Two possible optimal paths were built using correlation coefficients and, alternatively, their associated pvalues. The need for models created using high-throughput biological experiments is clear in the fight against Alzheimer’s disease. Its characterization has been very elusive. It is believed that optimization models, such as the TSP can support this effort in a transparent and effective way. Future work will consist on comparing multiple microarray databases at a time and to gather more biological evidence to evaluate the resulting configurations elucidated in this work.
Funder Acknowledgement(s): This material is based upon work supported by the National Science Foundation (NSF) under Grant HRD 0833112 (CREST program), as well as the National Institutes of Health (NIH) MARC Grant 5T36GM095335-02 “Bioinformatics Programs at Minority Institutions.”
Faculty Advisor: Mauricio Cabrera-Rios,