Menu
0 Comments

Machine learning will help to abandon animal testing

Machine learning will help to abandon animal testingChimpanzee Enos before placing the spacecraft mercury Atlas 5, 1961.

American scientists have managed to develop a computer method of detecting toxicity of chemical substances, superior in precision experiments on animals. To create it used the binary coefficients of similarity , and the system is trained on more than 866 thousand hazardous properties common in experiments chemicals. The article was published in the journal Toxicological Sciences.

Toxicological experiments conducted by most of the pharmacological, cosmetic and other companies that produce chemical products: they help to evaluate the safety of substances for human use. Despite the fact that such studies are regulated by ethical norms of individual institutions, the testing of drugs for toxicity on animals is widely condemned. Worldwide companies are trying to limit the use of animals in the stages of creating their products; however, the majority of undertaken measures are not effective enough: for example, in 2017 the number of laboratory animals used has increased by seven percent compared to the previous year.

Machine learning will help to abandon animal testingChimpanzee Enos before placing the spacecraft mercury Atlas 5, 1961.

American scientists have managed to develop a computer method of detecting toxicity of chemical substances, superior in precision experiments on animals. To create it used the binary coefficients of similarity , and the system is trained on more than 866 thousand hazardous properties common in experiments chemicals. The article was published in the journal Toxicological Sciences.

Toxicological experiments conducted by most of the pharmacological, cosmetic and other companies that produce chemical products: they help to evaluate the safety of substances for human use. Despite the fact that such studies are regulated by ethical norms of individual institutions, the testing of drugs for toxicity on animals is widely condemned. Worldwide companies are trying to limit the use of animals in the stages of creating their products; however, the majority of undertaken measures are not effective enough: for example, in 2017 the number of laboratory animals used has increased by seven percent compared to the previous year.

To help limit the number of animals used in Toxicological experiments can modern technology. It took a team of researchers led by Professor Thomas Hartung (Hartung Yhomas) from Johns Hopkins University. They used data on 80 thousand chemicals from the files collected by the European chemicals Agency: end dataset contains information about 866 thousands of properties and guidelines for toxicity and possible harmful effects for all analyzed chemicals. All data were obtained during long-term experiments on animals.

The trained system evaluates the likelihood of the hazardous properties of chemicals (toxicity upon contact, for example, lacrimal glands) based on the calculation of similarity coefficients: this phase of learning occurs without the teacher. Based on the data then, using the method of k-nearest neighbors, similarity properties of the evaluated chemicals is presented in the form of two-dimensional vectors. Next, the system was trained to assess the toxicity of the input data using the random forest algorithm that uses many decision trees.

READ  Minister of Communications and Mass Communications does not preclude the blocking of Viber

The researchers were able to achieve the accuracy of Toxicological analysis in 87 percent of cases: for comparison, a single experience involving animal has an accuracy of 57 percent and the re — 81 cent.

The data collected, therefore, was quite in order to teach the computer to evaluate the possible negative impact of chemicals to humans. The use of such methods, according to researchers, is a great opportunity to limit harsh treatment of animals in pharmacology and cosmetology.

Often animals are used not only in studies of the safety of substances and medicines, but also for behavioral studies. For example, last year scientists managed to get mouse models of bipolar disorder.

Tags:

Leave a Reply

Your email address will not be published. Required fields are marked *