Proceedings - Interdisciplinary Center for Nanotoxicity
Proceedings - Interdisciplinary Center for Nanotoxicity
Proceedings - Interdisciplinary Center for Nanotoxicity
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
136<br />
Conference on Current Trends in Computational Chemistry 2009<br />
Predictive QSAR modeling of animal toxicity endpoints using<br />
a combination of chemical and biological descriptors of<br />
molecules<br />
Alexander Tropsha<br />
Laboratory <strong>for</strong> Molecular Modeling, Carolina <strong>Center</strong> <strong>for</strong> Computational Toxicology<br />
and <strong>Center</strong> <strong>for</strong> Environmental Bioin<strong>for</strong>matics,<br />
University of North Carolina at Chapel Hill, Chapel Hill, NC U.S.A.<br />
Modern “‐omics” technologies produce a wealth of biological assay data <strong>for</strong> large<br />
numbers of chemicals with known in vivo toxicity profiles. For instance, to develop effective<br />
means <strong>for</strong> rapid toxicity evaluation of environmental chemicals, the Tox21 partnership between<br />
the National Toxicology Program (NTP), NIH Chemical Genomics <strong>Center</strong>, and the National<br />
<strong>Center</strong> <strong>for</strong> Computational Toxicology at the US EPA are conducting quantitative high‐<br />
throughput screening studies with thousands of chemicals. Concurrently, the ToxCast program<br />
established at the US EPA is addressing the Tox21 goals by using ca. 600 in vitro assays to create<br />
bioactivity profiles <strong>for</strong> 100s of compounds with known in vivo toxicities measured in ca. 70<br />
animal assays. The analysis of this data requires new computational approaches to link<br />
chemical structures, in vitro responses and in vivo toxicity effects.<br />
We have developed a general purpose predictive QSAR modeling workflow that relies on<br />
effective statistical model validation routines. Traditionally, we have used this workflow <strong>for</strong><br />
conventional QSAR modeling studies of in vivo and points utilizing chemical descriptors of<br />
molecules. Recently, we have started to employ both chemical and biological (i.e., in vitro assay<br />
results) descriptors of molecules to develop in vivo chemical toxicity models. We have<br />
developed two distinct methodologies <strong>for</strong> in vivo toxicity prediction utilizing both chemical and<br />
biological descriptors. In the first approach, we employ biological descriptors in combination<br />
with chemical descriptors to build models. Obviously, this approach requires the knowledge of<br />
biological descriptors to make toxicity assessment <strong>for</strong> new compounds. Our second modeling<br />
approach employs the explicit relationships between in vitro and in vivo data as part of the two‐<br />
step hierarchical modeling strategy. In the first step, all compounds are partitioned into classes<br />
defined by patterns of in vitro – in vivo relationships (e.g., compounds designated as “active” in<br />
both in vitro and in vivo assays <strong>for</strong>m one class whereas those that have these designations<br />
disagree <strong>for</strong> in vitro and in vivo assays, <strong>for</strong>m another class). Then, a conventional classification<br />
QSAR model using chemical descriptors only is built to discriminate compounds that belong to<br />
different classes defined by the in vitro – in vivo correlation patterns. In the second step, the<br />
class‐specific conventional QSAR models are built using chemical descriptors only. Thus, this<br />
hierarchical strategy af<strong>for</strong>ds external predictions using chemical descriptors only.<br />
We will present the results of applying the above QSAR modeling strategies to several<br />
datasets including ToxCast Phase I data; the ZEBET dataset with in vitro IC50 cytotoxicity values<br />
and in vivo rodent LD50 values <strong>for</strong> more than 300 chemicals; and others. All studies suggest<br />
that utilizing in vitro assay results as biological descriptors af<strong>for</strong>d prediction accuracy that is<br />
superior to both the conventional QSAR modeling that utilizes chemical descriptors only and<br />
the in vivo effect classifiers based on in vitro biological descriptors only. We will discuss how<br />
our models are being used to prioritize compound selection <strong>for</strong> the ToxCast Phase II studies.