Public Release: 

Insilico Medicine announces validation of the GAN-RL systems for de novo molecular design

InSilico Medicine, Inc.


IMAGE: Insilico Medicine announces experimental validation of the GAN-RL systems for de novo molecular design. view more 

Credit: Insilico Medicine

12th of July, 2018, Rockville, MD -- Today Insilico Medicine, a company specializing in artificial intelligence for biomarker and drug discovery and aging research, announced the experimental validation of the GAN-RL systems for de novo molecular design. Insilico applied its AI-powered algorithms to design new preclinical agents with desired properties, including novel inhibitors of JAK, BRAF, and EGFR kinases. The synthesis and biological evaluation of compounds was performed by a highly reputable validation partner.

Initially introduced by Ian Goodfellow, Generative Adversarial Networks (GANs) have demonstrated great potential in the generation of various types of real-world data, including images, texts, and video. Insilico Medicine subsequently pioneered the application of GANs in the field of drug discovery and has published numerous scientific papers on GAN architectures utilizing innovative and advanced approaches, including reinforcement learning (RL) and condition-based generation to generate novel molecular entities with pre-specified chemical and biological properties.

"The results of the biological testing are very promising. Such proof-of-concept studies lay the foundation for further implementations of artificial intelligence to drug discovery and allow us to look beyond the traditional paradigm. We are happy to share some of these results with the technology companies working on entering the pharma space during the ICML conference", said Alex Aliper, president of EMEA, Insilico Medicine.

The pharmaceutical companies attending the ICML 2018 conference in Sweden are encouraged to visit the booth #B07:13 near the poster area for a demonstration.

While the company is pursuing a large number of targets discovered using the multi-omics analysis in a variety of age-related diseases, compounds with high therapeutic potential in anti-cancer therapy, such as those targeting JAK, BRAF and EGFR kinases, were selected for synthesis and validation of the GAN-RL systems performance.

In the in vitro testing, generated compounds have demonstrated two-digit nanomolar activity against JAK kinases, including series of selective JAK-1 and JAK-3 inhibitors. Other molecules have shown potent inhibitory activities against BRAF and EGFR kinases. Condition-based GAN, turned on the activity data of known ligands, was able to generate dual BRAF-EGFR inhibitors with picomolar potency. Examined by highly experienced medicinal chemists, the compounds were marked as novel molecules with clear IP status and interesting structural motifs.

The model developed by Insilico Medicine fixes theoretical flaws in the initially proposed conditional version of the Adversarial Autoencoder, used in many previous papers by multiple researchers. The proposed model was intensively tested in multiple drug discovery-related tasks. The model extrapolated patterns of binding affinity to generate compounds with higher affinity to an MCL1 protein compared to molecules from the training data. Tuned on a small set of active and inactive molecules against the family of JAK kinases, the model was able to generate selective molecules against JAK-1 and JAK-3. Incorporating several reward functions in RL-based model allowed it to achieve greater structural diversity and shift the distribution of generated samples to more complex, nature-like chemistry.

"Our goal is to create an end-to-end AI-based pipeline that will allow to reduce the time and costs at all stages of drug discovery and deliver real cures to help millions of patients worldwide. We are excited to announce the results of the experimental validation of our GAN-RL engine. This approach has a vast potential to generate novel pre-clinical candidates. This engine is one of the many vital parts of our drug discovery pipeline", said Alex Zhavoronkov, Ph.D., founder, and CEO of Insilico Medicine.


For further information, images or interviews, please contact:
Contact: Qingsong Zhu, PhD

About Insilico Medicine, Inc

Insilico Medicine is an artificial intelligence company headquartered at in Rockville, with R&D and management resources in Belgium, Russia, UK, Taiwan, and Korea sourced through hackathons and competitions. The company and its scientists are dedicated to extending human productive longevity and transforming every step of the drug discovery and drug development process through excellence in biomarker discovery, drug development, digital medicine and aging research.

Insilico pioneered the applications of the generative adversarial networks (GANs) and reinforcement learning for generation of novel molecular structures for the diseases with a known target and with no known targets. In addition to working collaborations with the large pharmaceutical companies, the company is pursuing internal drug discovery programs in cancer, dermatological diseases, fibrosis, Parkinson's Disease, Alzheimer's Disease, ALS, diabetes, sarcopenia, and aging. Through a partnership with, the company launched a range of nutraceutical products compounded using the advanced bioinformatics techniques and deep learning approaches. It also provides a range of consumer-facing applications including Young.AI.

In 2017, NVIDIA selected Insilico Medicine as one of the Top 5 AI companies in its potential for social impact. In 2018, the company was named one of the global top 100 AI companies by CB Insights. In 2018 it received the Frost & Sullivan 2018 North American Artificial Intelligence for Aging Research and Drug Development Award accompanied with the industry brief. Brief company video:

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.