Professor Yang Shengyong's team published an article in Nature Communications to reveal a molecular generative model based on deep learning, and found that RIPK1 inhibitors have been studied...

Recently, the team of Professor Shengyong Yang of the Biotherapeutic Research Center of our hospital published an article in Nature Communications, a sub-journal of the journal Nature, revealing the research process of discovering RIPK1 small molecule inhibitors based on molecular generative models based on deep learning, showing the ability of generative deep learning (GDL) models to generate new molecular structures, highlighting the great potential
of deep learning in drug discovery 。 Postdoctoral Fellow Li Yueshan, Dr.
Zhang Liting, Dr.
Wang Yifei and Associate Professor Zou Jun of the State Key Laboratory of Biotherapy of Sichuan University are the co-first authors of the paper, and Professor Yang Shengyong is the corresponding author
of the paper.

In the early stage of drug development, how to efficiently discover emerging compounds or lead compounds with novel skeleton structures is a key step in innovative drug development, and it is also a very challenging task
.
The traditional strategy has been to utilize high-throughput screening methods to screen
from existing compound libraries.
However, due to the limited structural diversity of the existing compound library and the repeated screening of major pharmaceutical companies and drug discovery institutions, it has become increasingly difficult
to find active compounds with new skeleton structures with independent intellectual property rights.
Generative models based on deep learning can generate compounds with new backbone structures, which provides a new idea
for solving the dilemma encountered by traditional high-throughput screening methods.

In recent years, the research of generative models or generative deep learning (GDL) models based on deep learning has developed
rapidly.
Among them, GDL models based on recurrent neural networks (RNNs) are the most widely reported
.
The conditional RNN (cRNN) can explicitly guide the subsequent molecular generation process
by giving the initial state vector of the RNN as a condition.
However, existing cRNNs and other GDL models still have many shortcomings, such as over-reliance on objective functions, generative molecular novelty, and limited diversity
.
In addition, although most GDL models have been validated at a theoretical level, there are still few
examples of successful application to practical innovative drug discovery.

To this end, Professor Yang Shengyong's team proposed a new cRNN molecular generation model, which integrates transfer learning, regularization enhancement and sampling enhancement
.
The team further applied this model to the discovery
of RIPK1 kinase inhibitors.
By establishing a customized RIPK1 inhibitor virtual compound library, virtual screening, chemical synthesis and biological activity validation, a highly active and selective RIPK1 inhibitor (RI-962)
was obtained without any modification.
The team then resolved the crystal structure of RIPK1–RI-962, structurally elucidating the molecular mechanism
by which RI-962 has high activity and selectivity 。 The team also evaluated the in vivo effects of RI-962 on mouse models of TNFα-induced systemic inflammatory response syndrome (SIRS) and DSS-induced inflammatory bowel disease (IBD).
The results showed that RI-962 improved TNFα-induced SIRS and DSS-induced IBD damage
by inhibiting RIPK1 kinase activity.

Based on this, the research team proposed a new GDL model and used the model to discover a highly active and selective RIPK1 inhibitor
.
This study demonstrates the ability of GDL models to generate entirely new molecular structures, highlighting the great potential
of deep learning for drug discovery.

The research work has been strongly supported
by the National Natural Science Foundation of China Innovation Research Group Project, Key Project, General Project, Youth Science Foundation Project, West China Hospital 1.
3.
5 Project and National Postdoctoral Innovation Talent Program.

Article link: