Designing drugs with code: are we here yet?

author | Dr. Melissa Landon, Chief Strategy officer

CADD The field can be traced back to 70 The emergence of personal computers in the late s .1981 year ,CADD On the 《Fortunate》 The cover of the magazine , Known as the “ The next industrial revolution ”.

In the twinkling of an eye, forty years have passed , We have seen unprecedented levels of private investment in AI driven life sciences companies , And is building the next generation of artificial intelligence to assist drug discovery （AIDD） Platform technology company （ for example Recursion、Exscientia） And more traditional CADD company （ for example Schrödinger） A successful public appearance . Although it can be said , since CADD Since the early days , The field has matured considerably , Now includes AI Driven Technology , But the image Kate The scientists who laid the foundation decades ago are responsible .

The current pandemic has created an urgent need for the rapid development of therapies , This is a challenge that the pharmaceutical industry has been trying to deal with for decades .In silico The method is to reduce cost and time 、 Improve to develop design more efficiently - manufacture - test - analysis (DMTA) Cycle definition is promising for traditionally developed manual and expensive laboratory centric processes . We are beginning to see evidence that this commitment may be fulfilled ,Recursion company 、Exscientia Company and Insilico Medicine The company recently announced , These projects are entering the stage of clinical development , The time required is only a small part of the general time . Perhaps the most striking thing is , We are not only seeing a shift in the way drugs are designed , And also saw the location of the drug design . A recent study published by Boston Consulting emphasizes that the discovery pipeline of AI drug discovery companies is growing almost exponentially , It is in sharp contrast to the stagnation of the pipeline in the early stage of large pharmaceutical companies . More and more CADD and AIDD The company holds the problem in its own hands , Use their platform to create assets , Instead of simply authorizing their work to big pharmaceutical and biotechnology companies .

therefore , In all recent developments , The biggest problem is ： Have we reached ？ What is hype and reality ？ Can we shake the machine and make drugs ？ In my foolish opinion , The answer is No . There are many reasons , Mainly （1） Biology is very complex , Our understanding of the relationship between biology and disease is limited （2） Many components of the drug discovery process are not optimized for speed and scale , as well as （3） We still lack enough data of high quality and quantity , In order to make full use of the potential of artificial intelligence .Andreas Bender and Isidro Cortés-Ciriano An opinion published last year summed it up well [2]." In short , Artificial intelligence in drug discovery requires meaningful quantitative variables and labels , But we often don't have enough ability to determine which variables are important , There is not enough ability to define these variables experimentally （ And large enough ）, Nor is there enough capacity to label the success of artificial intelligence with a biological label commensurate with the current investment and hope in this field ." let me put it another way , although （ so to speak ） Conventional CADD The method has reached " Productivity Plateau " Stage , But artificial intelligence in drug discovery is still in the hype cycle " The peak of exaggerated expectations " Stage .

But reasons for optimism abound . automation 、 Advances in scalable and reliable chemical synthesis and experimental testing have brought great hope to solve some challenges related to data generation and data quality . just as Peter Diamondis In his 《 The future is faster than you think 》 As asserted in a Book , The integration of technology, artificial intelligence and human intelligence will be a necessary condition for realizing a great leap . Before that , The impact of silicon-based methods on drug discovery is still limited . However , As a hopeful skeptic , I am still encouraged , As an industry , One day we will use code to influence the design of all drugs . then , We will go back to " dangerous " On board .

Reference material

• Jayatunga M et al. AI in small molecule drug discovery: a coming wave? Nat Rev Drug Discov. 2022. 21: 175-176
• Bender A and Cortés-Ciriano I. Artificial intelligence in drug discovery: what is realistic, what are illusions? Part 1: Ways to make an impact, and why we are not there yet. Drug Discov Today. 2021. 26(2): 511-524.