Maxtox – predicting toxicity online ! Part 1

What is toxicity ? This question has been asked by a newbie to biology, as well as the forty year old drug research scientist. The newbie wants to know the meaning of this word, while the scientist wants to know why is it being used to shelve the drug that he had researched for the last decade ! Indeed this word brings doom to a new drug, it can wipe out companies and research worth many million dollars can come to naught.

The thrust therefore now (and for a long time now) has been to understand this term in the clearest way possible before beginning the long and winding road to drug discovery. There are tomes of books and research materials trying to help in this direction, but when it comes to the practicalities – toxicity is a really difficult concept to understand and measure. There are a variety of “standard” experiments and there are a multitude of competing definitions. Today, the new drug discoverer could not be in a more turbulent cauldron of information, definitions and experiments. Even the experiments themselves are long drawn out affairs consisting of many steps and a lot of resource investments.

In recent times a part of this problem has been solved by software tools. These run on HPCs (High Performance Clusters) and serve a variety of purposes in the aid of drug discovery. A few purposes that these tools serve are

1. Help in understanding the structure and 3d spatial orientation of the new molecule.
2. Help in understanding the activity of the molecule towards a target protein etc.
3. Determine the toxicity of the molecule by comparing it to various standards and formulas ( and other molecules)

However a couple of  limitations of these tools has always been the processor intensive nature of the calculations as well as the cost if installing the software and the hardware required for running these systems. The build up to the actual predictive numbers was a major hurdle for most fledgling drug companies. Enter Opentox.

So what is Opentox ? Briefly its a worldwide project (funded under FP7 of EU Commission) to get predictive technologies onto the web, where they can be easily accessed by scientists and interested individuals. A part of this project which is being implemented by me is called MaxTox. Maxtox is a predictive algorithm which tries to answer the important question – how toxic is this new molecule ?

Before running into what Maxtox does, lets first understand a little more about this word – toxicity. There a few common definitions abounding on the internet. Here are a few important ones :

• TD₅₀ : Chronic dose rate (expressed in milligrams per kilogram of bodyweight per day) which would induce tumors in half of the number of test animals at the end of the standard lifespan for the specie.
• LD₅₀ :  The LD₅₀ is the dose that kills half (50%) of the animals tested (LD = “lethal dose”). The animals are usually rats or mice, although rabbits, guinea pigs, hamsters, and so on are sometimes used.

As we can see there a few important things in these definitions. The first one is the dose rate – expressed in milligrams per kilogram of bodyweight per day. The second is the endpoint – the target species or the organ of a target species. For example the liver of a rat. There could be a series of experiments done on the liver of live rats to quantify the toxicity of a chemical.

In the toxicity definition, you should be able to notice another important thing – although only in a implied form. Its the number of animals that have to be put to death to get these values. This however is the bitter truth about drug discovery – you have to test it in a multitude of animals before it is considered safe for humans. However Maxtox can help alleviate this problem to a great extent.

Ok so lets find out more about Maxtox. Here are a few things that would define this algorithm.

1. Maxtox predicts the toxicity of a chemical against a particular end point (like rat liver TD50).
2. Maxtox requires existing data of toxicity against the predicted endpoint – preferably a large set of molecule already tested.
3. Maxtox compares the new molecules structure to the tested molecules to find similarities.
4. Maxtox uses these similarities in structure to make a statistical model – which predicts the toxicity.

Heres a link to try out the service : http://opentox2.informatik.uni-freiburg.de:8080/MaxtoxTest/. Right now there is one model in place so you can get a prediction of only one type of toxicity. Later the intended direction is to have a multitude of models.

In later articles we I will discuss advantages of this algorithm and the internal logic.

No related posts.