Opportunities and challenges facing AI and drug development

The development of new drugs is characterized by high cost, long development cycle and low success rate. According to Nature, the new drug R&D cost is about 2.6 billion US dollars, which takes about 10 years and the success rate is less than 1/10. How to speed up the development of new drugs and reduce the R&D cost has become an urgent task for major pharmaceutical companies.

Recent advances in artificial intelligence techniques for learning and predicting new features, especially deep neural networks (DNNs) or recurrent neural networks (RNNs), have led to the widespread use of artificial intelligence technology and the acceleration of social automation. Against such backdrop, the application of artificial intelligence technology combined with big data and cloud computing in drug R&D is increasing, and the application advantages are also highlighted.

Application scenarios and technologies of combing AI and drug development

Since the Dartmouth Conference in 1956, AI has been used in drug discovery for more than 60 years. It has now penetrated into all stages of pharmaceutical R&D, but it has also focused on the discovery and validation of new drugs. However, the applied technology has made great progress, and the training data sets and models from the previous quantitative relationship (QSAR) and quantitative structure-property relationship (QSPR) research have progressed to machine learning, cognitive computing and image recognition.

Now, the main scenarios of AI and drug development include: exploring drug targets, mining drug candidates, high-throughput screening, drug design, drug synthesis, predicting drug ADMET properties, pathophysiological studies, and development of new indications using old drugs. Target screening is the most popular field of the c model, and the combination of the two will bring new drugs to new heights, while small molecule drug screening and design still dominate.

At present, representative start-ups for “AI+ drug development” include Exscientia, Benevolent AI, Atomwise, Relay Therapeutics, Numerate, IBM Waston and Lam Therapeutics. According to the research focus of existing start-ups in the therapeutic field, tumors account for the largest proportion, while the neurological field is second, and there are more rare diseases.

Advantages of “AI+ drug development”

Compared with the traditional drug R&D model, the new “AI+ drug R&D” model has the advantages of shortened time cycle, less capital costs and improved success rate by making full use of existing medical resources. According to statistics, it may take 4-5 years for the drug development in the traditional mode to be in the preclinical stage. The new drug development pipeline based on AI and bio-computing can complete pre-clinical drug development on average 1-2 years, and drug development is significantly accelerated. Since then, the first drug-turbocharged flu vaccine designed entirely through AI has entered the clinical phase.

Opportunities and challenges facing AI and drug development

A report from TechEmergence shows that artificial intelligence can increase the success rate of new drug development from 12% to 14%, saving the biopharmaceutical industry billions of dollars. In addition, it has been reported that AI can save 40%-50% of the time in compound synthesis and screening compared to traditional means, saving the drug companies $26 billion in compound screening costs per year. In the clinical research phase, it can save 50%-60% of the time, saving $28 billion in clinical trial costs per year. That is to say, AI can save $54 billion in R&D costs for pharmaceutical companies every year. Compared with the traditional model, AI+ drug development has obvious time and cost advantages. The future market of “AI+ medicine” has great potentials. By 2025, the market for “AI+ drug R&D” will exceed $3.7 billion (excluding medical treatment, etc.).

But this model also faces many challenges. In April 2019, IBM decided to stop developing and selling drug development tools—the Watson Artificial Intelligence Suite, because of its poor financial performance. As a leader in artificial intelligence in the field of medicine and health, it has to face a state of financial downturn. In addition, the current AI application is more focused on target screening, and has now screened many targets through literature analysis, but the confirmation of the target is a difficult problem. Last but not least, the prediction of the drug's drug-making properties by AI is lower than that obtained through trial and analysis.

Therefore, on the whole, the real output of “AI+ drug research” is very small. Therefore, companies need to rationally position their role in the industry chain and choose the appropriate innovative business model.

In addition, companies that develop drugs based on AI also face challenges from policies, talents, and technology. The introduction of new technologies will change the original drug research and development model, and the regulatory talents and policy guidelines need to be updated simultaneously. However, there are no targeted policy guidelines. In terms of talent, the lack of high-end talent also limits the development of this field. In addition, the understanding of AI drug development awareness and biological complexity also needs to be improved. In the data issues that determine the quality of AI+ drug research and development, how to establish R & D data standard system to improve the data, how to establish a risk-sharing mechanism, is also what the future AI+ drug research and development needs to face.

Conclusion

Although currently the “AI+ drug R&D” model still faces many challenges, it is clear that this combination is bound to be the future development trend of the pharmaceutical industry in the next ten or even twenty years. BOC Sciences is a highly proved provider of comprehensive drug discovery services, which includes various screening libraries like Fragment Library, Activity based Libraries and Custom Libraries as well as services like Hit to Lead, Lead Optimization, Chemical Resynthesis, drug testing service, building block synthesis, synthesis service, hit identification service, and more.