Predictive systems toxicology

Waters, M., and Jackson, M. (2008). Databases applicable to quantitative hazard/risk assessment— Towards a predictive systems toxicology. Toxicol Appl Pharmacol 233, 34-44. 

As described in Chapter 9 there are an increasing number of commercial toxicological prediction systems available. Naturally these have been designed to be user friendly most run under Microsoft Windows and use the Simplified Molecular Input Line Entry System (SMILES) as the molecular input. It is therefore possible to obtain a prediction of toxicity instantaneously, and often this may be performed for large numbers of compounds. There is a great temptation to use predicted toxicities at face value (i.e., if a compound is predicted to be non-toxic then it must be non-toxic). This simplistic use of predicted values should be avoided at all costs. Ideally, there are a number of criteria that should be applied when predicting toxicity. It is essential that a trained expert uses the predictive system. The user should be an expert both in the endpoint being predicted and the use of the predictive system. 

Richard AM. Commercial toxicology prediction systems a regulatory perspective. Toxicol Lett 1998 102-103 611. 

Ridings JE, Barratt MD, Cary R, Earnshaw CG, Eggington CE, Ellis MK, Judson PN, Langowski JJ, Marchant CA, Payne MP, Watson WP, Yih TD. Computer prediction of possible toxic action from chemical structure an update on the DEREK system. Toxicology 1996 106 267-79. 

Koski WS, Kaufman JJ. TOX-MATCH/PHARM-MATCH prediction of toxicological and pharmacological features by using optimal substructure coding and retrieval systems. Anal Chim Acta 1988 210(l) 203-7. 

This raises the key test of target validation. Does a compound that has the appropriate potency, selectivity, and ADME properties, that is active in "predictive" animal models, and is free of other systems toxicology work in the targeted human disease state This is the ultimate test of a target-based drug discovery program. With the considerable compound attrition rates in moving from animals to diseased humans, this has not proven to be a predictable transition. A case in point is that of the NKl receptor activated by the peptide, substance P. 

Purpose The dermal absorption data enable EPA to make risk assessments when the oral or inhalation route determined the toxic effects in defined faboratory animal models yet the exposure to humans is expected by the dermal route. A complete kinetic analysis is essential to convert oral or inhalation low-effect and no-effect doses into dermal low-effect and no-effect doses. This affows for the calculation of margin of exposure or risk to predict systemic toxic effects that otherwise may not be tested practicalfy by the dermaf route. The rat is the preferred model because a large toxicology database exists in this species. The rat absorption... 

The DEREK program is an expert system that attempts to provide expert level guidance in evaluating and predicting the toxicological properties of small (<64 atoms) molecules. 

The consequence of moving consciously toward this model will be the provision of a robust and scalable IT infrastructure and systems able to cope with exponentially growing data mountains that will need to be integrated and shared, accessed and mined in the most effective way. It will also require formidable computing power and sophisticated algorithms to be able to simulate both organs and whole body systems to reduce expensive failures in the clinic and predict much earlier the pharmacokinetic and pharmacodynamic properties and toxicological and efficacy profiles of molecules in pharmaceu-... 

These techniques may be coarsely classified into techniques that mimic human reasoning about toxicological phenomena (Expert Systems) and methods that derive predictions from a training set of experimentally determined data (Data Driven Systems). 

This approach is intuitively appealing to most users, because it promises easy access to toxicological knowledge, and some of the most used predictive toxicology software tools are in fact Expert Systems [e.g., Derek Nexus (https //www. lhasalimited.org/) and Toxtree (http //toxtree.sourceforge.net/)]. 

There have been several attempts over the years to classify adverse drug reactions, primarily based on the perception of what can and what cannot be predicted given knowledge of the pharmacology and toxicology of the drug. For example, Edwards and Aronson [13] proposed what is probably the most extensive classification system ... 

---