AI powers drug R&D OpenAI technology promises a new technological revolution

With the development of technology, AI technology has been widely used in several medical field scenarios. According to GlobAI Market Insight, the global healthcare AI market will be US$ 4.2 billion in 2020 and is expected to hit US$ 34.5 billion by 2027. In terms of the share of each segment, drug R&D is the top-ranked segment.

Examples of AI applications in the drug R&D process

One of the typical representatives of AI applications in the field of drug R&D is the development of the AI AlphaFold system by DeepMind. By continuously training and iterating AlphaFold using a database of nearly 170,000 different protein structures and a database of protein sequences containing unknown structures, the AlphaFold system learns amino acid sequences and has the ability to accurately predict protein structures.

In November 2020, AlphaFold 2 predicted most of the protein structures in CASP 14, a protein structure prediction competition, by only one atom's width from the real structure, reaching the level of human observation and prediction using sophisticated instruments such as cryoelectron microscopy, which is unprecedented and great progress in protein structure prediction.

In China, there are already many enterprises dedicated to the application of AI in the field of new drug screening, such as the DELT compound library and virtual compound library built by Pharmastone Technology, the DEL coded compound library, SBDD and FBDD screening libraries built by Hitgen Inc. and Viva Biotech respectively.

Traditional drug screening methods are huge, time-consuming and inefficient. High-throughput screening technologies that combine AI with these drug databases can narrow the search area faster and more accurately, freeing drug developers from a lot of repetitive work, shortening the time to develop new drugs, reducing costs, and improving R&D efficiency.

Capital Pursuit Hot Areas

Starting in March, TandemAI announced the completion of a US$ 35 million Series A financing led by Qiming Venture Partners and followed by OrbiMed, Eight Roads, and F-Prime Capital. Founded at the end of 2021, TandemAI is committed to building a new model of drug discovery with computation at its core. This financing will be used mainly for the refinement of the drug discovery service platform, which aims to integrate physical modeling, and AI with interdisciplinary cross-applications in biophysics, medicinal chemistry, and biology

In addition to such startups, AI medical enterprises that have been deeply involved in the industry for several years have developed to Series C, D and E and have ushered in an IPO boom in 2021, with Keya Medical, Airdoc Technology, InferRead, and Shukun Technology submitting their prospectuses.

According to incomplete statistics, throughout 2022, the global AI drug R&D-related financing events amounted to 144, with a total amount of more than US$ 6.2 billion, the total amount of financing increased by 47.6% compared to 2021, which is the hot area of pharmaceutical R&D in recent years.

The application of AI in the drug R&D as represented by OpenAI technology

Recently, ChatGPT exploded in a short time and reached 100 million users in two months. ChatGPT is a large language model for modeling and predicting natural language input, capable of answering human questions and conducting natural conversations. This type of AI is gradually playing a role in drug R&D, in addition to its current search and chat function.

In late January, a paper published in Nature Biotechnology revealed that ProGen, a new AI system developed by Salesforce (a software service provider), is capable of generating artificial enzymes from scratch.

ProGen is an "OpenAI"-like semantic recognition system in which Salesforce scientists fed amino acid sequences of 280 million different proteins into a machine learning model. By learning what amino acids used to be in a given original sequence, the new ProGen model can create protein structures that are not found in nature with simple instructions.

In comparison, while traditional protein drug design relies on random amino acid mutations and optimization of natural proteins, Progen's model revolutionizes the design process of such protein-based drugs and enables deep learning to create new protein sequences that are different from most natural proteins, which is a disruptive innovation and advancement.

As mentioned above, AlphaFold 2, a next-generation of AI technology for drug R&D, will enable accurate prediction of drug structures, while ProGen, a next-generation of OpenAI technology, will directly simulate the new structures of synthetic proteins, which will greatly expand the initial protein structure library for drug R&D and bring a new technological revolution to current drug R&D.