An Assessment of ILP-Assisted Models for
Toxicology and the PTE-3 Experiment 

Ashwin Srinivasan1, Ross D. King2 and Douglas W. Bristol3

1 Oxford University Comp. Lab., Wolfson Bldg., Parks Rd, Oxford, UK
2 Dept. of Comp. Sc., University of Wales Aberystwyth, Ceredigion, UK
3 NIEHS, Lab. of Carcinogenesis and Mutagenesis, RTP, NC, USA



Abstract. The Predictive Toxicology Evaluation (or PTE) Challenge
provided Machine Learning techniques with the opportunity to compete
against specialised techniques for toxicology prediction. Toxicity models
that used findings from ILP programs have performed creditably in the
PTE-2 experiment proposed under this challenge. We report here on an
assessment of such models along scales of: (1) quantitative performance,
in comparison to models developed with expert collaboration; and (2)
potential explanatory value for toxicology. Results appear to suggest the
following: (a) across of range of class distributions and error costs, some
explicit models constructed with ILP-assistance appear closer to optimal 
than most expert-assisted ones. Given the paucity of test-data, this
is to be interpreted cautiously; (b) a combined use of propositional and
ILP techniques appears to yield models that contain unusual combinations 
of structural and biological features; and (c) significant effort was
required to interpret the output, strongly indicating the need to invest
greater effort in transforming the output into a toxicologist-friendly
form. Based on the lessons learnt from these results, we propose a new
predictive toxicology evaluation experiment  PTE-3  which will address
some important shortcomings of the previous study.
References

1.	J. Asliby. Predictions of rodent carcinogenicity for 30 compounds. Environmental
Health Perspectives, pages 11011104, 1996.
2.	R. Benigni. (Q)sar prediction of chemical carcinogenicity and the biological side
of the structure activity relationship. In Proceedings of The Eighth International
Workshop on QSARs in the Environmental Sciences, 1998. Held in Baltimore,
May 1620, 1998.
3.	J. Bootman. Speculations on the carcinogenicity of 30 chemicals currently under
review in rat and mouse bioassays organised by the us national toxicology program.
Mutagenesis, 27:237243, 1996.
4.	L. Breiman, J.H. Friedman, R.A. Olshen, and C.J. Stone. Classification and Regression 
Trees. Wadsworth, Belmont, 1984.
5.	D.W. Bristol, J.T. Wachsman, and A. Greenwell. The NIEHS Predictive-Toxicology 
Evaluation Project. Environmental Health Perspectives, pages 1001
1010, 1996. Supplement 3.
6.	L. Dehaspe, H. Toivonen, and R.D. King. Finding frequent substructures in chemical 
compounds. In Proceedings of the Fourth International Conference on Knowledge 
Discovery and Data Mining (KDD-98), pages 3036. AAAI Press, 1998.
7.	J. Huff, E. Weisburger, and V.A. Fung. Multicomponent criteria for predicting
carcinogenicity: dataset of 30 ntp chemicals. Environmental Health Perspectives,
104:11051112, 1996.
8.	J.E. Huff, J.K. Haseman, and D.P. Rail. Scientific concepts, value and significance
of chemical carcinogenesis studies. Ann Rev Pharmacol Toxicol, 31:621652, 1991.
9.	D.F.V. Lewis, C. Ioannides, and D.V. Parke. COMPACT and molecular structure
in toxicity assessment: a prospective evaluation of 30 chemicals currently being
tested for rodent carcinogenicity by the NCI/NTP. Environmental Health Per-
spectives, pages 10111016, 1996.
10.	C.A. Marchant. Prediction of rodent carcinogencity using the DEREK system for
30 chemicals currently being tested by the National Toxicology Program. Environmental 
Health Perspectives, pages 10651074, 1996.
11.	F. Provost and T. Fawcett. Analysis and visualization of classifier performance:
comparison under imprecise class and cost distributions. In Proceedings of the
Third International Conference on Knowledge Discovery and Data Mining (KDD-
97), pages 4348. AAAI Press, 1998.
12.	F. Provost and T. Fawcett. Robust classification systems for imprecise environments. 
In Proceedings of the Fifteenth National Conference on Artificial Intelligence 
(AAAI-98). AAAI Press, 1998.
13.	R. Purdy. A mechanism-mediated model for carcinogenicity: model content a
prediction of the outcome of rodent carcinogencity bioassays currently being conducted 
on 25 organic chemicals. Environmental Health Perspectives, pages 1085
1094, 1996.
14.	R.Benigni, C. Andreoli, and R.Zito. Prediction of the carcinogenicity of further
30 chemicals bioassayed by the US National Toxicology Program. Environmental
Health Perspectives, pages 10411044, 1996.
15.	A. Srinivasan. Note on the location of optimal classifiers in n-dimensional ROC
space. Technical Report PRG-TR-2-99, Oxford University Computing Laboratory,
Oxford, 1999.
16.	A. Srinivasan and R.D. King. Using Inductive Logic Programming to construct
Structure-Activity Relationships. In Proceedings of the AAAI Spring Symposium
on Predictive Toxicology. AAAI Press, Menlo Park, CA, 1999. (to appear).
17.	A. Srinivasan, R.D. King, and D.W. Bristol. An assessment of submissions made
to the Predictive Toxicology Evaluation Challenge. In Proceedings of the Sixteenth
International Conference on Artificial Intelligence (IJCAI-99,). Morgan Kaufmann,
Los Angeles, CA, 1999. (to appear).
18.	A. Srinivasan, R.D. King, S.H. Muggleton, and M.J.E. Sternberg. The Predictive
Toxicology Evaluation Challenge. In Proceedings of the Fifteenth International
Conference on Artificial Intelligence (IJCAI-97,). Morgan Kaufmann, Los Angeles,
CA, 1997.
19.	R.W. Tennant and J. Spalding. Predictions for the outcome of rodent carcinogenicity 
bioassays: identification of trans-species carcinogens and non-carcinogens.
Environmental Health Perspectives, pages 10951100, 1996.
20.	Y.T. Woo, D.Y. Lai, J.C. Arcos, M.F. Argus, M.C. Cimino, S. DeVito, and
L. Keifer. Mechanism-based structure-activity relationship (sar) analysis of 30
ntp test chemicals. Environ. Carcino. Ecotox. Revs. C, 15:139160, 1997.
