CAPCODRE: A Computational Systems Biology and Machine Learning-Based Approach to Predict Cognitive Disorder Risk in the Elderly

Overview

As global life expectancy improves, the population of the elderly, persons that are aged 65 years and older, is steadily increasing as well. However, with aging populations a greater prevalence of cognitive impairment has emerged, ranging from mild dementia to severe dementia such as Alzheimer’s disease due to genetic and environmental influences, among others. The purpose of this research was to develop a computational algorithm to predict the risk of developing cognitive disorders using a dual machine learning and systems biology approach. The proposed method CAPCODRE (Computational Approach to Predict Cognitive Disorder Risk for the Elderly) utilized air, water, and noise pollution data coupled with a gene-protein interaction network, in addition to cognitive impairment hospitalizations in the United States to create a tailorable, interactive network able to predict risk of dementia and Alzheimer’s disease. This network was inputted into a random selection optimization algorithm that selected optimal training parameters for training via KNN, random forest regression, and decision trees. CAPCODRE was successfully able to predict and model risk of cognitive health issues through measures of specificity, sensitivity, and accuracy of >90%. The algorithm was integrated into an app for users to receive personalized predictions based on their medical history and geographic location. CAPCODRE can point to the extent of the effect of environmental pollution on human health and reveal steps to mitigate risk of severe cognitive impairment. This research also has the potential to address racial disparities in cognitive disorder diagnoses and treatment, promoting more equitable and accessible care. PUERTO RICO Statistical Assessment of TESS Demographics and Prospect Earth-Twins Via Habitability Constraints in Python Emily N. Alemán García CROEC, Ceiba, Puerto Rico Supervising Mentor: Nicholas Lauersdorf Considering the increasing amounts of planetary data available, this investigation aimed to automate the process of identifying likely habitable exoplanet candidates within big datasets-based on planetary radii, ESI, and HZ location. The researcher developed a computational model in Python to digest TESS data, survey selected because of short orbital period candidates suitable for atmospheric characterization. The model-was proven to be successful for (1) mining prospective rocky exoplanets within the HZ, (2) analyzing demographic tendencies- throughout filtering stages, and (3) incorporating post-processing visualization. TESS data tends towards those candidates likely incapable of harboring water due to bias in detection method by instrument sensitivity. The researcher hypothesized that likely habitable exoplanets share constant qualities that may influence radii, stellar flux, and HZ. Large-exoplanets tend to be detected orbiting around K, G, F, A-stars, while Earth-twins and Super-Earth’s are detected orbiting M, K, Gstars. Luminosity and stellar flux within the HZ remain within constant ranges of .004-4.441 and .213-1.758, respectively. These values coincide with the luminosity and stellar flux ranges of M, K, Gstars, spectral types hosting majority of exoplanets which survived all constraints, thus supporting my hypothesis. The-researcher created a TESS Habitable Exoplanet Catalog (THEC) to narrow the candidates for spectroscopy and follow-up confirmation. A THEC candidate was referred to the list of ACWG priorities of the TESS Follow-Up Program (TFOP).

Competition history

JSHS 2023 Category not listed

Resources

National JSHS abstract book

Related projects

ISEF · 2023