Dedication: 20-30h per week (partially remote)
Where: ETSIINF – UPM Campus de Montegancedo
Scholarship: 700-1000€ per month (depending on dedication and CV)
How to apply: Candidates should send an email to david.perez.rey@upm.es with CV
and academic records (grades)
AI-based New Approach Methodologies on Chemical Toxicity
Prof. David Pérez del Rey (in collaboration with ISCIII)
Current chemical hazard assessment relies heavily on traditional animal studies, which
provide dose-response data commonly summarized as NOAELs or BMDs. However,
these endpoints alone do not capture the full complexity and uncertainty inherent in the
original study data. New Approach Methodologies (NAMs), including in vitro and in silico
methods, offer mechanistic insights and high-throughput bioactivity data, but their
regulatory acceptance remains limited partly due to uncertainties in hazard
characterization and lack of direct comparability with traditional data.
The objective of this internship is to move beyond summary metrics to richer quantitative
data. Parallelly, comprehensive bioactivity datasets from the EPA’s CompTox Chemistry
Dashboard (including ToxCast and Tox21assays) will be integrated. This initiative
represents a paradigm shift in regulatory toxicology.
Specific Objectives:

1. Extract and curate bioactivity profiles from scientific literature, databases (e.g.
   CompTox), and available NAM-based assessments (e.g. OECD IATA cases
   studies) for relevant chemical substances.
2. Develop and validate explainable AI-driven pipelines trained specifically for
   automated extraction of endpoints and dose-response data from in vivo toxicity
   studies focusing on those considered valid by regulatory (e,g, EFSA and ECHA)
   assessment.
   Methodology:

• Retrieve, standardize, and extract bioactivity data from scientific literature,
  available case studies and CompTox Chemistry Dashboard including ToxCast
  and Tox21 assay results.
• Use state-of-the-art NLP techniques and locally trained LLMs to extract dose-
  response data points and metadata from textual and tabular content of toxicity
  endpoints described in in vivo studies validated in EFSA and ECHA dossiers