The Human Brain Pharmacome - An Overview.pdf (3.07 MB)

The Human Brain Pharmacome: An Overview

Download (3.07 MB)
posted on 19.04.2018, 20:56 by Charles Tapley Hoyt, 0000-0002-9820-6925, 0000-0002-2046-6145, 0000-0003-4153-3930, 0000-0001-9012-6720

The quantity of data, information, and knowledge in the biomedical domain is increasing at an unprecedented rate — with no signs of deceleration. Even with the assistance of information retrieval technologies, it is overwhelming, if not impossible, for individuals or groups of researchers to be knowledgeable of the state-of-the-art in any but an incredibly specific topic. Extracting and formalizing knowledge in a computable form allows computers to assist researchers in reasoning and analysis in a reproducible manner on a much larger scale.

Previous knowledge assemblies in the context of Alzheimer's disease have used the Biological Expression Language (BEL) to qualitatively describe the differences between healthy and diseased states [1] and identify underlying mechanistic signatures that are more representative of the pathophysiology of the complex disease than individual targets [2]. The Human Brain Pharmacome (HBP) will build upon previous annotations of select drugs' mechanisms of action [3] by using BEL as a semantic framework to systematically integrate quantitative pharmacokinetic and pharmacodynamic (PK/PD) information from publicly available, structured databases. The introduction of quantitative information to previously qualitative mechanistic knowledge assemblies is the first step towards enabling mechanistically-driven chemoinformatics.

Because knowledge-driven approaches to model the relevant biology and chemistry are inherently limited by the completeness and correctness of their associated knowledge assemblies, two complementary automated update techniques are being developed. First, natural language processing will be used to continuously extract biomedical relations from the recent literature and prioritize for semi-automated curation [4]. Second, deep-learning techniques will be used to extract quantitative PK/PD information directly from tables embedded in the literature.

The HBP will be used to develop several orthologous in-silico techniques for proposing repurposing candidates then validating them experimentally in-situ and in-vitro. The updating, enrichment, and experimental procedures will first be applied to a prioritized set of mechanisms underlying Alzheimer's disease based on their information density, assay-ability, and novelty because of the time restriction on curation and developing assays. The results from the experimental systems will be fed back into the HBP and used to further iterate through the knowledge discovery process.