Photo: Shutterstock

Click for chemistry

Thursday 24 Nov 16
by Karoline Lawætz


Eva Bay Wedebye
Senior Officer
National Food Institute
+45 35 88 76 04
A few mouse clicks—and several microseconds later—you have access to a world of data on more than 600,000 chemical structures.

Companies wanting to use new chemicals in their products and requiring a thorough knowledge of the substances’ properties prior to product launch can now breathe a sigh of relief. The same applies to researchers and authorities, for example, whose work involves approving chemical substances, as the new Danish QSAR database—the first of its kind in the world—contains finished analyses of the substances for online access. The database is located on DTU Food’s website. Here, over the past decade, a huge server has been packed with massive volumes of data and advanced software.

The result is a lightning-fast chemical database containing unprecedented volumes of information on the properties of more than 600,000 chemical substances.

Danish Environmental Protection Agency uses it on a daily basis

Magnus Løfstedt, Deputy Head of Division at the Danish Environmental Protection Agency, says that he uses the Danish QSAR database almost daily.

“A chemical name doesn’t tell you much about substance toxicity—here QSAR is a great help, as it saves us a lot of time. However, if you are planning to use the results in connection with substance regulation, for example, you will need further data backed up by expert assessment."

It began with tadpoles

"A model is based on existing measurements. We are trying to find measurements which tell us precisely what we need to know about a chemical substance"
Eva Bay Wedebye, chief adviser at DTU Food

The first QSAR models were developed more than a century ago. Here, researchers tested anaesthetic on tadpoles to identify a correlation between the effect of the substances and their chemical properties.

It turned out that the more lipophilic the substances were—i.e. the better they mixed with fats—the better their anaesthetic properties. The results were recorded using logarithmic paper and pencil.

Fortunately, modern QSAR methods and powerful computers eliminate the need to use tadpoles as laboratory animals. The system is able to predict how a substance will behave in a trial environment—without prior testing in the laboratory.

The better QSAR analyses—the fewer laboratory animals needed. On the other hand, it requires top-flight mathematical models.

Photo: Shutterstock

Just like the weather forecast

The mathematical models in the Danish QSAR database can predict whether a substance will be harmful or not and—among other things—it is the easy access to these analyses which makes the system unique.

“A model is based on existing measurements, just like a weather forecast. We are trying to find measurements which tell us precisely what we need to know about a chemical substance,” says Eva Bay Wedebye, Chief Adviser at DTU Food.

A model is based on a training set of an effect—e.g. whether substances are oestrogen-like in laboratory tests. If, for example, you have 500 known substances that have been tested in a particular protocol in the laboratory, you can create a model predicting the effect of unknown substances based on the results of these tests.

The goal is to identify the chemical properties that make some of the 500 known substances oestrogen-like—and others not. The known substances are analysed in a multitude of ways, enabling the computer to distinguish between problematic and non-problematic chemical compounds.

To enhance analytic credibility, the Danish QSAR database displays the results from three different types of software so users can see whether the various models produce the same result. The system also provides a general prediction based on the results from all three systems.

The laboratory tests—which form the starting point for QSAR—are always subject to a degree of uncertainty. The same is true of the mathematical modelling, which is why the analyses are not 100 per cent reliable.

“For certain purposes, QSAR analyses can’t stand alone, but they can be valuable as a planning tool or in supplement laboratory testing,” says Eva Bay Wedebye.

Not limited to experts

The database is free of charge, but you need some knowledge of chemistry to benefit from the analyses.

QSAR (Quantitative Structure-Activity Relationship)

QSAR models are computer models that can predict the properties of chemical substances. Based on experimental results from laboratory tests, the models have been developed by experts with knowledge of chemistry, biology, toxicology, statistics, mathematics, and computer science.

The Danish QSAR database contains predictable data on more than 600,000 organic ‘small molecules’—small chemical compounds with a carbon skeleton—e.g. physical-chemical properties, environmental impact, and toxic effects.

Pros and cons
There are other QSAR databases, but they do not function in the same way. Here, each substances is entered separately—even structural information—in order to perform QSAR model calculations. These systems are designed exclusively for experts.

In the Danish model, all the predictions are made in advance so that when you search for a chemical and an analysis, the program retrieves the results immediately, providing a comprehensive overview. However, the database is only useful if it contains the chemical in question.

The interface is easily accessible and developed so it works equally well for experts and others with an interest in chemistry—e.g. students or NGOs.

A QSAR analysis can be done as early as in the early development phase of new drugs, for example, before a chemical substance has even been manufactured in the laboratory and tests have been performed at the cellular level—in vitro—or with laboratory animals—in vivo. The models save time, money, and laboratory animals.

The analyses will always be prone to a certain degree of uncertainty in relation to experimental tests. That said, you can combine knowledge from models for a wide range of different qualities, thereby improving accuracy.

Try QSAR here:
4 APRIL 2020