You are here:
RESPONSE TO "THE OPTIMAL FRAGMENTATION PRINCIPLE".
Citation:
Richard, A M. RESPONSE TO "THE OPTIMAL FRAGMENTATION PRINCIPLE". DRUG DISCOVERY TODAY 6(5):235-237, (2001).
Description:
Response to "The Optimal Fragmentation Principle"
To follow up on my comment to Dale Johnson and to respond to his entertaining and provocative letter, I confess to some frustration in seeing a show-and-tell of a sophisticated tool for data mining and exploration, without seeing how the tool will realistically benefit the research community working in the toxicity modeling area. The fact is that these new and costly data mining tools require large stores of data relative to a defined endpoint, and are being marketed to and will primarily benefit the data richest among us -- i.e. the pharmaceutical and chemical industries.
In spite of the many factors in common in the approaches, structure-based toxicity prediction historically has been a poor cousin of drug design. As opposed to highly focused efforts to optimize for a particular and often well understood biomolecular interaction, toxicity prediction attempts to anticipate one or more potential adverse health effects, given many possible mechanisms by which such effects can occur, and with little understanding or data relative to many of these mechanisms. It is widely acknowledged, and regularly lamented within the toxicity modeling community, that access to quality toxicity data for sufficient numbers and diversity of chemicals, across the spectrum of toxicity endpoints of potential interest, is THE major factor limiting development of useful structure-activity models and expert toxicity prediction systems. The fact is, the larger and more diverse the training set, the better the toxicity prediction model. For example, a model that incorporates little information on pharmaceuticals is unlikely to accurately predict toxicity for pharmaceuticals.