Public Release: 

Achieving high-value chemical diversity for the pharmaceutical artificial intelligence

druGAN: A concept of a Generative Adversarial Autoencoder de Novo Generation of New Molecules with Desired Molecular Properties

InSilico Medicine, Inc.


IMAGE: Insilico Medicine Pharma.AI drug discovery pipeline utilizing GANs and reinforcement learning generates the high-value lead molecules view more 

Credit: Insilico Medicine, Inc.

Friday, October 13th, 2017, Baltimore, Maryland: Insilico Medicine announces the publication of a new research paper in Molecular Pharmaceutics titled: "druGAN: An Advanced Generative Adversarial Autoencoder Model for de Novo Generation of New Molecules with Desired Molecular Properties in Silico." Modern Generative Adversarial Networks (GAN)s achieved unprecedented accuracy and quality in image, video and text generation. The fundamental principle of GANs is adversarial training based on game theory results. The competition between the Generative and Discriminative networks leads to joint evolution and almost perfect results.

One of the most significant tasks at Insilico Medicine is adapting best neural network architectures for drug discovery process and it is committed to publishing the proof of concept advances that are at least one year old. These advances are usually integrated into a comprehensive drug discovery pipeline with the goal to enable the deep neural networks to produce perfect molecules for the specific set of diseases.

DruGAN allows the generation of new formulations for a wide range of diseases: different cancers, neurodegenerative diseases such as Alzheimer's disease, virus infections, and more. Of course, DruGAN is not a silver bullet and for successful usage; it requires a large team of professionals in both AI and medicinal chemistry. One of the limitations of the published approach is the use of the binary molecular fingerprints and the need to match the output molecules to the chemical libraries. To overcome these barriers, Insilico Medicine transitioned to novel representations of molecular structure based on the molecular graphs and presented the work at its annual "Artificial Intelligence and Blockchain for Healthcare" forum in Basel, Switzerland in September.

"Insilico Medicine has a policy of publishing the proof of concept research, which is one year or older to attract more data scientists to work on the healthcare problems. DruGAN is one of these proofs of concept. Internally the company switched to GANs with reinforcement learning (RL), which is essentially the environment that rewards GANs for generating "effective" novel molecular graphs. The molecules discovered using these techniques went through the in vitro validation and are undergoing in vivo testing. The use of GANs with RL is likely to transform the pharmaceutical industry", said Alex Zhavoronkov, Ph.D., founder and CEO of Insilico Medicine, Inc.



druGAN: An Advanced Generative Adversarial Autoencoder Model for de Novo Generation of New Molecules with Desired Molecular Properties in Silico

About Insilico Medicine, Inc

Insilico Medicine, Inc. is an artificial intelligence company located at the Emerging Technology Centers at the Johns Hopkins University Eastern campus in Baltimore, with R&D resources in Belgium, Russia, and the UK sourced through hackathons and competitions. The company utilizes advances in genomics, big-data analysis, and deep learning for in silico drug discovery and drug repurposing for aging and age-related diseases. The company is pursuing internal drug discovery programs in cancer, Parkinson's Disease, Alzheimer's Disease, ALS, diabetes, sarcopenia, and aging. Through its Pharma.AI division, Insilico provides advanced machine learning services to biotechnology, pharmaceutical, and skin care companies, foundations and national governments globally. In 2017, NVIDIA selected Insilico Medicine as one of the Top 5 AI companies in its potential for social impact. Brief company video:

Contact: Qingsong Zhu, PhD

Alex Zhavoronkov, PhD

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.