Learning rules for comparing the activities of Alzheimer's drugs
Contents and links
Introduction: Alzheimer's disease
Alzheimer's disease has long been recognised as one of the most disabling
conditions affecting the aged. It is the major cause of dementia among
elderly people. The drug Tacrine has drawn considerable attention because
of a report of its clinical efficacy for this problem. Its structure is
However, large-scale clinical trials show that the primary disadvantage
of Tacrine is its high toxicity. This has led to investigations into possible
variants of the basic Tacrine structure. Some reported variations used
the following structure
In this structure, new compounds are created by substituting for R and
X, where the substitution for X can be in position 6 or 7.
The need to guide experimentation
The literature contains many reports of tests on the properties of new
drugs - in the case of Alzheimer's, desirable proprties include
However, these reports do not indicate any particular pattern followed
in creating these new drugs. The aim of our research is to learn rules
that tell us what makes a drug have a desirable property (for example,
low toxicity). Hopefully, these rules will offer better guidance for the
synthesis of new drugs.
High acetocholinesterase inhibition.
Good reversal of scopolamine induced deficiency.
Inhibit amine re-uptake.
Using ILP to learn drug properties
In our experiments, data from the chemical literature was converted into
a representation suitable for ILP, as described in [King,
R.D., Srinivasan, A. and Sternberg, M.J.E.(1995)]. The chemical data
itself is taken from [Shutske G.M. et al(1989)].
In the dataset provided here, rules are to be learnt for each property
using the following information:
The positive and negative examples are pairwise comparisons of drugs. There
are different sets of examples for each property.
Physical and chemical properties of chemicals that can be substituted for
R or X. These are described by unary predicates in the background knowledge.
These properties include hydrophobicity, hydrophilicity, charge, size,
polarity, whether a chemical is aliphatic or aromatic, whether it is a
hydrogen donor or acceptor etc.
Sizes, hydrophobicities, polarities etc., are represented by constants
such as polar0 and polar1.
Relations between the constant, expressed as clauses such as less_than(polar0,polar1),
are also provided as background knowledge.
x_subst(Drug,Position,Subs). The X substitution for Drug
in position Position is Subs.
The R substitution is split into parts corresponding to bonds between different
Sometimes the R substitution involves the presence of 1 or more ring structures.
These are benzyl derivatives that arise from substitutions within the basic
benzene ring structure. The number of such substitutions in the basic benzene
structure, the position of the substitutions an the actual substituent
The Golem dataset
The data used in these experiments is contained in one
compressed TAR file.
King, R.D., Srinivasan, A. and Sternberg,
Relating chemical activity to structure: an examination of ILP successes.
New Gen. Comput. (to appear).
Shutske G.M., Pierrat F.A., Kapples
K.J., Cornfeldt M.L., Szewczak M.R., Huger F.P., Bores G.M., Haroutunian
V. and Davis K.L. (1989).
9-Amino-1,2,3,4-tetrahydroacridin-1-ols: Synthesis and Evaluation
as Potential Alzheimer's Disease Therapeutics.
Up to applications main page.