New Horizons in Toxicity Prediction.
Lhasa Limited Symposium Event in Collaboration with
the University of Cambridge - February 2009
A Report by Wendy A. Warr, wendy@warr.com, http://www.warr.com Question
from the floor: Will you spend more time on biologicals now?
Boyer:
Biologicals cannot be used in all areas. We must run studies in species
closer to humans. The ambiguity of exposure is another issue.
Matthews:
This is a significant gap: 40% of our information is on biologicals. We
need expert system rules for these, for example peptides with
immunologic toxicity.
Question
submitted in writing: Is rodent bioassay for carcinogens useful?
Benigni:
Yes. All human carcinogens, when adequately tested, were correctly
identified in rodents. You cannot test potential carcinogens on humans;
anyway it would take 30 years.
Richard:
Multispecies hits are of the greatest significance, so we must test in
multiple species, and multiple cancer sites, to get significant results.
Benigni:
In addition, there is a correlation between rat and mouse, and between
animals and humans in terms of carcinogenic potency. This is a strong
support for extrapolation from rodent bioassay to humans. On the
contrary, the results in terms of target organs are quite idiosyncratic.
Matthews:
With new pesticides there is a data gap in our understanding of
pancreas, thyroid, etc. variation among animals. We need to get to
grips with this.
Question from the floor: Concerning
carcinogenicity across species and sexes: how many carcinogens have not
shown up in shorter studies in animals? They should be detectable in
shorter studies.
Matthews:
Expert systems were very successful; 90-day and six month studies can
be useful.
Benigni:
There is no relationship between short and long term. Long term rodent
positives and negatives are reflected in humans.
Glen:
In identifying flu epidemics Google searches have been monitored. Could
we do this with adverse drug reactions? How about social networks in
toxicology?
Boyer:
We could also look at people’s electronic patient records.
Richard:
Going back to an earlier point, the huge variation in individuals, look
at Jeremy Nicholson’s work on the metabolome. We can look at
general
trends and gross generalizations. It is interesting how much you can
explain by what people eat. Nicholson can see changes before there is
any histopathology.
Guengerich:
In a study of two sets of rats it turned out that an environmental
factor (i.e., which room the rats were in) could be involved. Think
also about transcriptonomics.
Matthews:
There are so many chemicals for which we have no toxicology data. Of
160,000-180,000 common chemicals we have toxicology on only 5%. Let us
find the really bad ones: it could be a huge success in a short time.
Alan
Wilson: Pick out the really bad molecules in pharma.
From
the floor:
We need data curation and ontologies to put the data in the right
format. In metabolomics this is being done. Apply the tools to safety
as well as to efficacy.
Boyer:
The experimentalists will change the way they do experiments when they
see how successful models are.
Question
from the floor: Should we be suspicious of 70% specificity?
Matthews:
We can model some toxicities (e.g., endocrine, heart and kidney) but
the liver is harder.
Question
from the floor: Will there be a balance in future of in silico versus
experimental?
Matthews:
The National Cancer Institute (NCI) uses a battery of cell lines and
they are extraordinarily successful. QSAR could not do that. But at the
next stage we will want to know adverse effects of the hits etc. so you
will need both methods in future.
Hirose:
Look at the overall success of Ames tests. In silico gives
some
information sometimes. We can use it on a case by case basis.
Cronin:
This is not a case of one solution fitting all situations. There are
roles for both in silico
and in vitro.
Hence the importance
of ITS.
Richard:
You will always need in
silico where you do not have the chemicals (for
example, virtual libraries), but we want to use both methods.
Guengerich:
You might ask whether organic chemists at Cambridge should spend all
their time in the laboratory or all their time in the library.
Greene:
Experimental data from a well validated biological assay should always
be considered to be more reliable than an in silico
prediction.
Richard:
I disagree: it depends.
Matthews:
in picking the first dose for clinical trials you set up a QSAR and you
will get very near, but the result from complex animal analysis is
worse. Animals have different bioavailability and different metabolism.
Benigni:
The Ames test is one of the best in
vitro assays, but results vary from
laboratory to laboratory.
From
the Floor: Carcinogenicity prediction works within an
application domain.
Greene:
Even with one receptor the chemistry “space” being
synthesized changes
over time and so an in
silico model fails to predict for new compounds
coming through.
Question
from the Floor: What about metabolism based toxicology?
Richard:
Some of the ToxCast in vitro assays will have metabolic capability.
Guengerich:
For the FDA you have to do the experiments to find the metabolites, to
get round expensive retesting.
Matthews:
QSAR tools can be used as a method of prioritizing. They can cope with
the most likely off-target activities for metabolites as well as for
the original chemicals. You confirm your in silico
observations with
wet work.
Boyer:
You
start with a molecule of molecular weight 300 or 400. You scale back
the functionality to make the molecule specific but when it is
metabolized, more functional groups appear, and the metabolite is
therefore less specific.
From
the floor: Pharma is making molecules live longer in the
body. Now we need to predict the effect of the drug on the environment.
David
Hawkins: Are there examples?
Benigni:
There is no experience in Europe.
Matthews:
In pre-manufacture notice in the United States for food you have to use
in
silico methods.
Richard:
Pre-manufacture notification in the Toxics program requires EPA to make
a toxicity determination without data; hence, SAR and in silico methods
have been necessary and essential to this program. In the pesticide
program, in contrast, EPA has had legal authority to require lots of
test data; hence, SAR historically has not been used. Congress has
recently mandated, however, that pesticides programs evaluate
tolerances for impurities in pesticides, without giving them the
authority to request new data; hence, the pesticides program now has to
look to SAR and in
silico methods.
Matthews:
Tier testing can detect the highly toxic compounds so you can make
early decisions on some pesticides.
