Scalable Biomarker Discovery for Diverse High-Dimensional Phenotypes

Abstract

Historically, most biological and medical investigations have examined a few discrete outcomes of interest, and only a few controllable parameters were modified to perturb or improve these outcomes. Investigations of this form went hand-in-hand with the development of inferential statistics, which provide the quantitative tools to detect which perturbations successfully improve outcomes. Biological and clinical research has entered a realm of modifying hundreds or thousands of experimental parameters in high throughput, however, and high dimensional statistics have been developed to understand which of these modifications in turn significantly improve outcome. The field has now reached a point where hundreds or thousands of outcomes can be simultaneously measured as well, but few statistical tools exist to answer the question, "When many experimental or patient outcomes are measured simultaneously, and many experimental parameters or treatments are modified, which modifications are significantly associated with improved outcomes?" This project thus aims to develop novel statistical methods for efficiently associating many controllable predictor variables with many observed response variables with high sensitivity and specificity.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 23, 2015
Accession Number
AD1007191

Entities

People

  • Curtis Huttenhower

Organizations

  • Harvard T.H. Chan School of Public Health

Tags

DTIC Thesaurus Topics

  • Bile
  • Biostatistics
  • Colitis
  • Computational Biology
  • Computational Science
  • Computer Science
  • Data Sets
  • Engineering
  • Engineers
  • Gene Expression
  • Genetics
  • Gut Microbiome
  • Microbiomes
  • Microorganisms
  • Public Health
  • Statistics
  • Students

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Distributed Systems and Data Platform Development
  • Systems Analysis and Design