Opportunities and challenges in the development of AI-plus drugs

The research and development of new drugs has three high-risk properties with high cost, long research and development cycle and low success rateThe cost of developing new drugs is about $2.6 billion, takes about 10 years, and has a success rate of less than one-third, Natrue reportedHow to speed up the research and development process of new drugs and reduce the cost of research and development has become an urgent problem for major pharmaceutical companiesIn addition, changes in drug distribution and the healthcare value chain have forced pharmaceutical companies to lower prices and increase the value of their drugsthe cumulative data for drug research and development is growing at a rapid rate, and the digital transformation of drug research and development is acceleratingTherefore, the first task of pharmaceutical companies is to use this data to drive value, to achieve the ultimate goal of increasing drug productivity and approval rates, and reducing costsrecent advances in artificial intelligence technologies to learn and predict new features, especially deep neural networks (DNNs) or recursive neural networks (RNNs), have made a wider range of applications and increased social automationUnder this background, artificial intelligence technology combined with big data and cloud computing is increasingly used in drug research and development, and the application advantages are also highlightedAI-plus drug research and development application scenario and technologyfrom the 1956 Dartmouth Conference, AI has been used in drug research and development for more than 60 years, has now penetrated into all stages of pharmaceutical research and development, but also mainly concentrated in the new drug discovery and verification stageHowever, the applied technology has made great progress, from the previous quantitative relationship (QSAR) and quantitative structure-relationship (QSPR) research mark-training data sets and models to machine learning, cognitive computing and image recognitionNow, the main scenarios for a combination of AI and drug development include: discovering drug targets, excavating drug candidates, high-throughput screening, drug design, drug synthesis, predicting the properties of drug ADMET, pathophysiology research, and the development of new indications - new use of old medicinesAmong them, target screening is the most popular field of recent AI-plus drug research and development, and the combination of the two applications will also enable the new use of old drugs to reach new heights, but small molecule drug screening and design still occupy a major positionBut at the speed of the application scenario, drug synthesis may become the most automated direction in the futureThe AI techniques commonly used in these applications are mainly machine learning, cognitive computing and image recognitionAI-plus drug development represents the enterprise and layout sector
And currently, a representative start-ups in AI-plus drug development include Exscientia, Benevolent AI, Atomwise, Relay Therapeutics, Crystal Technologies, Numerate and IBM Waston and Lamlams According to the existing start-ups in the field of treatment layout, the proportion of tumors, and the field of neurology, and rare disease-related enterprises are also more Therefore, the oncology and nervous system is not only the current aI-plus drug development layout focus area, but also the potential area for future development, and AI will also help to crack the rare disease diagnosis and drug development difficult "dilemma" situation Figure 1 The layout of the treatment field of a drug research and development start-up data source: biopharmatrend, data cut-off time is the end of July 2019
AI-plus drug research and development advantages and representative examples
compared with the traditional drug research and development model, AI-plus drug research and development has the advantages of shortening the research and development cycle, saving capital costs, improving the success rate, making full use of existing medical resources and other advantages According to statistics, the traditional model of drug development alone is clinical stage may take 4-5 years The new drug research and development pipeline based on AI and biocomputing can complete preclinical drug research and development in an average of 1-2 years, and drug research and development has been significantly accelerated Since then, the first fully AI-designed drug, the turbo-turbo" flu vaccine, has entered clinical stage Pharnext's combination therapy PXT3003, developed by AI technology to treat fibula atrophy 1A subtype, has completed two Phase III clinical sessions with positive results In 2017, Tianshili also reached a cooperation agreement with Pharnext AI-plus drug research and development enterprise sits in cooperation On September 11, 2019, Jiangsu Howson and Atomwise announced that they will work together to design and discover up to 11 potential candidates for undisclosed target proteins in several therapeutic areas Under the agreement, the potential total value of the partnership for Atomwise will exceed $1.5 billion Previously, Zhengda Qing worked with Alibaba Cloud to obtain a new compound screening method It is reported that compared with traditional computer-aided drug design methods, this new method can improve screening accuracy by 20% look at the cooperation of global pharmaceutical companies, so far, the world's top ten multinational enterprises have entered the field of AI-plus drug research and development, the specific cooperation information is shown in the table below Therefore, in the field of aI-research drugs, start-ups lay out platforms that make money through technical cooperation, while traditional large enterprises enter the market through cooperation or strategic investment Under this trend, the emergence, co-operation and financing of start-ups reached an all-time high in 2018, with start-ups raising $1,036 million (not including undisclosed items) Source: Public Data Collating
Opportunities and challenges to ai-plus drug development a report from TechEmergence shows that artificial intelligence can increase the success rate of new drug development from 12 percent to 14 percent, saving the biopharmaceutical industry billions of dollars In addition, AI is reported to save 40 to 50 percent of its time on compound synthesis and screening than traditional methods, saving pharmaceutical companies $26 billion a year in compound screening costs During the clinical phase, 50-60% of the time can be saved and $28 billion annually in clinical trial costs AI could save pharmaceutical companies $54 billion a year in research and development costs Compared with the traditional model, the time and cost advantage of AI-plus drug development is obvious now, the world's top 10 pharmaceutical companies are in the market, and start-up financing and collaboration is at an all-time high, leading the way to high financing returns, and priority distribution of large companies such as Roche has mastered high-quality data sources According to this development, the future aI-plus medicine market has great potential for development By 2025, the market size of AI-plus drug development will exceed US$3.7 billion (excluding diagnosis and treatment, etc.) , but a myopharmaceutical sdevelopment also faces less optimistic status quo and many challenges In April 2019, IBM decided to stop developing and selling the Watson AI Suite, a drug development tool, because of poor financial performance As a leader in artificial intelligence in the field of pharmaceutical and health, we have also had to face the state of weak financial performance , the current AI application of more concentrated target screening direction, has been through literature analysis and other screening than the approved drug more targets, but the target is indeed a difficult problem, how to establish a corroboration model, and what to confirm, human and financial resources to keep up, this is also need to think In addition, artificial prediction of the drug's medicinal availability, compared with the drug obtained through testing, can be less convincing Because predictions are based on data sets of fewer than 2,000 approved drugs (not necessarily of high quality), this is far from the most basic requirement for deep learning that relies on high-quality, identifiable data sets This also happens to be the advantage of AI's application in drug synthesis therefore, on the whole, the real output of AI-plus drug research and development is very small, and most companies need to face the situation that the result of insufficient or inadequate output results and financial distress Therefore, enterprises need to reasonably locate the role of the industrial chain, choose the right innovative business model in addition , a company that develops and develops a drugs, also faces policy, talent, technology and other challenges The introduction of new technology, so that the original drug research and development model changes, regulatory personnel, policy guidelines, etc need to be updated simultaneously, and now there is no targeted policy guidelines As far as talent is concerned, the lack of high-end composite talent sits also limits the development of this field And the "black box" feature of AI multitasking is still the resistance of deep neural networks to extract key correlation information from complex biological information In the future, it is necessary to improve the policy supervision simultaneously, to train the complex high-end talents, and the technical aspects such as the practical development of natural language processing, the multi-dimensional application of knowledge map, and the knowledge question-and-answer, analytical decision-making and semantic search also need to be greatly improved In addition, the understanding of the cognitive and biological complexity of AI-plus drug development needs to be improved In determining the quality of aI-drug research and development data, how to establish a standard system of research and development data to improve data, how to establish a sharing mechanism of risk-interest, is also the future aI-drug research and development needs to face conclusion Although the current situation of AI-plus pharmaceutical research and development is not very optimistic, but also face many challenges, but it can be clear that the combination of AI-plus drug research and development is bound to be the future development trend of the pharmaceutical industry, but also in the next ten or even twenty years, the pharmaceutical field to carry out a disruptive revolution, ushering in a new era.