GenAI: Expediting drug discovery while reducing development costs

Raj Babu, Founder & CEO of Agilisium highlights that GenAI is transforming drug discovery by reducing time and costs. By leveraging advanced data analysis and predictive modelling, GenAI accelerates each stage, from target identification to clinical trials, enabling faster, cost-effective delivery of innovative therapies

Developing new and innovative drugs has always been a time-consuming and expensive endeavour. On average, it takes 12 to 15 years for a drug to go from molecule to market, with 7 to 10 years dedicated to preclinical research. Unfortunately, only one out of ten drugs that enter clinical trials are approved. Recent research by McKinsey indicates that it costs up to $2.8 billion to develop a new drug. 

While traditional drug discovery includes vast laboratory research, starting with target identification, validation, assay development, High Throughput Screening (HTS), lead optimisation, ADME properties and animal model efficacy, broad preclinical testing, and multiple phases of clinical assessment, it is complex and creates huge burdens in finances and time. 

Fast-paced technological changes are underway, nevertheless, to drive a sea change in the pharma industry. One of the most promising developments has to do with the use of generative AI in drug discovery. But how precisely is GenAI speeding up the drug discovery pipeline while cutting down on development costs? 

Before delving into Generative AI solutions for rapid drug development at a reduced cost, let’s examine the major challenges faced across the stages of drug discovery

Major challenges in traditional drug discovery 

The different stages of drug discovery are inherently challenging. At the stage of discovery itself, molecular modelling is complex, genomics and proteomics data are integrated inefficiently, and high-throughput screening is limited and often missing the most promising candidates. Moving into the preclinical phase, the selection of a viable drug candidate from thousands of compounds is enormously selective, especially when the human response is translated from animal models and overall drug safety is ascertained through toxicity studies. Biologics have less than a 10 per cent success rate in clinical trials, with costs reaching up to $310 million per trial. Realtime data analysis, adaptive trial designs, and patient stratification are all part of the bundle of challenges too. Generative AI addresses many of these challenges with solutions focused on better quality, accuracy, and efficiency, reimagining the drug discovery pipeline.

Challenges to solutions 

Now, let us see why generative AI solutions are changing the drug discovery pipeline. This is because newer developments such as decreasing costs of high-throughput sequencing technologies, whole genome sequencing, digitisation of health records, and increased computing capabilities have resulted in a deluge of biomedical data. These developments have placed GenAI as a potential solution for the effective analysis of these large datasets, hoping for quicker drug development using data-driven methods.

Role of GenAI in target identification, lead optimisation, and clinical trials Target identification: Scientists spend countless hours trawling through patents, scientific literature, and trial data just to work out diseases and potential drug targets. A gruelling task, often in large volumes, means it isn’t done thoroughly enough. LLM-powered knowledge extraction hands the tough duty of understanding the analysed text, images, and other types of data over to AI. In contrast to old methods using natural language processing, new AI tools like GenInsights can help delve deeper into medical contexts. This will help the researcher specifically in phases like lead optimisation and target identification by offering the opportunity to ask detailed questions and switch tasks with ease without the need for extensive training to customise information for special needs, adding more evidence in the process. While reducing the time taken to identify the target, GenAI is also likely to reduce the costs associated with iterative cycles of synthesis testing!

Lead optimisation: In the drug discovery process, once the target is identified and validated, the assay is developed, and millions of small molecules are screened through High Throughput Screening (HTS). From millions of molecules, a few thousand molecules are selected as leads. Lead-to-lead optimisation processes are made faster with advancements in AI-based drug discovery tools for drug binding, homology modelling, toxicity prediction, drug filters like the Lipinski rule, reactive molecule filtration, and PK-PD modelling. Also, with extensive data, these models develop deep insights into large and small-molecule chemistry. GenAI models help predict parts of molecules, such as atoms in small molecules or amino acids in larger ones.

This could potentially shorten preclinical testing timelines by 20 to 30 per cent, allowing hopeful candidates to transition to clinical trials more quickly.

Preclinical testing: The capability of knowledge extraction that empowers researchers to decide on the medical conditions or indications a particular molecule will target is one of the critical decisions in biopharma. Of course, traditional methods would run the risk of missing real key evidence. Pharma companies have started using GenAI in the preclinical phase to analyse Real-World Data (RWD) and Molecular Knowledge Graphs, to discover new insights in indication selection. RWD is an underappreciated resource for indication selection, and models analysing medical events as texts to show patient-related similarities. Molecular knowledge graphs can chart protein pathway relationships that yield new indications that were experimentally validated in preclinical models. 

Clinical trials: In clinical trials, where time is money and the most important resource, GenAI makes all the difference. It simulates trial scenarios and predicts compound efficacy and safety profiles, reducing the number of trials and shortening the time frame of regulatory approval. That’s expected to reduce the cost by up to $300 million per successful drug brought to market. Overall, GenAI provides perfect optimisation of each phase of drug development besides allowing researchers to find new areas of innovation, ultimately giving life-saving therapies to patients worldwide by increasing the speed, cost-efficiency, and success rate. While GenAI is revolutionising drug discovery, it is also empowering small pharma and research institutions to be more competitive s rate. While GenAI is revolutionising drug discovery, it is also empowering small pharma and research institutions to be more competitive. 

GenAI and its impact on small pharma and research institutions 

Generative AI now is well known, for its increasing contribution to new prescription drug approvals and new molecular entities (NME). The discovery of such NMEs by small pharma drastically surged from 31 per cent to 64 per cent in just ten years. Also, GenAI solutions enhance the quality and precision of drug development by 60–80 per cent, hence highlighting their centre stage in reshaping the pharma world. Now, let us look at use cases where they use GenAI. 

AI-driven protein design by small pharma companies: Small pharma companies are looking to leverage generative AI to help biologists design better proteins to realise their functions. These companies are using AI models running on huge datasets of protein sequences and lab-generated proprietary data to speed up the production of bio-based products required for human and environmental health. This could turn any research and development into a very high-paced process. 

Disease prediction models by research institutions: Research institutions in emerging countries like India have access to country-specific disease data. For example, cervical cancer or breast cancer-related statistics across multiple Indian states. These research institutions are converting their raw data into insights using GenAI and even going beyond – such as developing a disease prediction model app that can help doctors and patients identify potential comorbidities.

Clinical study reports (CSR) to research papers: Another area where research institutions and a few emerging pharma companies are leveraging GenAI is shortening the time taken to publish their research papers. They feed CSRs into their custom GenAI models and receive research papers (journal manuscripts) with 95 per cent accuracy, saving a significant amount of time and effort. Although we have many use cases that are enhancing the drug discovery process, we should also consider the ethical side of using GenAI.

Ethical considerations in drug discovery 

While AI is transforming the possibilities of drug discovery, ethical considerations arise, and responsible innovation must be upheld. These include mitigating bias through diverse datasets, ensuring transparency in algorithmic decision-making, protecting patient privacy, providing equal access to medicines, and operating within the strictest regulatory boundaries. By addressing these considerations, we can provide assurance and fairness in every medical advancement.

Conclusion 

Now is the time to reimagine the drug discovery strategy using the power of GenAI. This less-travelled road, like any new technology, will come with its challenges and unique opportunities. As discussed earlier, with the help of GenAI, we can accelerate drug discovery, reduce its cost, and enable swift treatments while upholding our commitment to ethical principles that prioritise patient well-being and scientific integrity. Pharma companies and research institutions, whether big or small, should take a systems-based approach when utilising GenAI, which means having a holistic view. This includes balancing investments, returns, risks, and the value delivered to stakeholders. Achieving this will not only require bold and innovative leadership but also a culture that always puts ‘humanity’ first.

AgilisiumAI-driven protein designClinical study reportsDisease prediction modelsGenAIHigh Throughput ScreeningMcKinseyRaja Babu
Comments (0)
Add Comment