An Automated Scientific Discovery Framework (ASDF) for Mechanistic Reasoning Across Complex Data

Abstract

During this reporting period the HMS team continued to develop an approach to systematically explain gene dependencies and identify characteristics of genes frequently involved in unexplained correlations and clarifying their function through curation using large network models assembled from machine reading. The HMS team also developed an entity disambiguation system to eliminate false-positive explanations, and a method to normalize sites of phosphorylation to improve interpretation of phosphoproteomic data. The University of Arizona team developed machine-learning based systems for improved relation polarity detection and biological context extraction, and made technical improvements to the software underlying REACH. The OHSU team continued to develop methods for evaluating genetic and regulatory predictors of cellular phenotypes and identifying underlying causal relationships that explain phosphoproteomic differences observed in leukemia samples.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 14, 2019
Source ID
W911NF1810124

Entities

People

  • Peter K. Sorger

Organizations

  • Army Contracting Command
  • Defense Advanced Research Projects Agency
  • Harvard Medical School

Tags

Readers

  • Molecular and genetic basis of cancer.
  • Neural Network Machine Learning.
  • Regression Analysis.

Technology Areas

  • AI & ML
  • AI & ML - Information Retrieval
  • Biotechnology