How AI can speed drug development

Visit our Life Sciences webpage

Development of a new drug often takes 10 to 15 years, costs billions of dollars-and fails close to 90% of the time, according to one recent study. Life sciences companies must do better if they are to thrive in an era of spiraling costs, increased competition, and pressure to deliver new treatments more quickly.

\r\n

Historically, drug development followed a linear path. Researchers would identify a target, screen existing compounds and optimize them to create new drugs. Artificial intelligence-in which advanced algorithms analyze vast data sets to suggest new solutions to problems-can collapse that linear path, blurring the sequential lines between discovery and analysis.

\r\n

Today's AI-driven drug discovery platforms do this by integrating various public, commercial and custom catalogs for virtual screening. Leveraging analysis and visualization capabilities and advanced features like generative AI-based models, these platforms effectively expedite the generation of novel molecules. The most important benefit, however, might just be how these AI platforms expand the accessibility of that data-extending it across all stakeholders for seamless collaboration.

\r\n

Small wonder, then, that one report found 84% of current users and 70% of current non-users expect AI to drive significant or even transformative impact in drug discovery over the next five years. Here are some of the areas in which that impact is (or will be) most apparent:

\r\n

1. \r\n
2. Automated discovery platforms
   \r\n
   \r\n The drug discovery process has evolved from manual methods to advanced automated techniques. Think of an AI-powered drug discovery platform as a "prime mover," or source of propulsion, that links and unleashes the combined power of the drug discovery ecosystem-providing insights, predictions and solutions that were previously challenging to obtain.
   \r\n
   \r\nModern computational strategies have significantly accelerated discovery processes. Target structure prediction, for example, anticipates the 3-D structure of proteins. Binding site analysis uses virtual fragment simulations to identify and characterize these crucial sites. In QSAR predictions, a chemical's physical structure is used to predict its biological activity or toxicity.
   \r\n
   \r\nIn the field of AI in drug development, these improvements have enabled precise prediction of pertinent physiochemical processes and ADMET (absorption, distribution, metabolism, excretion and toxicity) properties, which can reduce the drug failure rate due to toxicity and safety issues. In as little as 40 days, researchers can now identify, synthesize and test 20 candidates with a 10% success rate using AI, compared to 1% in the past.
   \r\n
   \r\n
3. \r\n
4. AI for predicting protein structure
   \r\n
   \r\n Other uses of AI include protein structure prediction using deep learning, which enables accurate 3D protein structures that aid drug design, and generating novel molecular structures through advanced AI techniques.
   \r\n
   \r\nProtein language models (PLMs) analyze protein sequences just as natural language processing (NLP) models process human language. They excel at tasks such as de novo protein sequence generation, controllable protein design, and protein property prediction.
   \r\n
   \r\nLike NLP models, PLMs are trained on vast datasets based on protein sequences that contain the instructions for creating proteins. The models learn from these sequences and gain insights into the underlying syntax of proteins to predict how different variations impact their function, stability and interactions. This enables both the virtual screening of compound libraries for drug candidates and the assessment of potential side effects on candidates based on their protein interactions.
   \r\n
   \r\nSparked by the commercial success of RNA vaccines for COVID-19, biotechnology firms are now pursuing therapeutics based on engineered circular RNA (circRNA). RNA, in its usual linear form, is short-lived. However, circRNA's increased stability could enhance its therapeutic potential, even at low-dose levels. AI-based methods such as DRfold have improved the accuracy of RNA models by more than 70%, which can aid in the design of RNA-targeted small molecules. It must be noted that while DRfold has great potential, it remains somewhat unproven.
   \r\n
   \r\nRNA molecules can adopt specific 3D motifs that are considered druggable, offering untapped potential to therapeutically modulate numerous cellular processes, including those linked to protein targets previously considered "undruggable."
   \r\n
   \r\n
5. \r\n
6. Literature search and knowledge management hubs 
   \r\n
   \r\n The influx of information during the literature review portion of drug discovery research can be overwhelming. AI-based tools can help by leveraging technologies such as text mining and NLP to automate key pieces.
   \r\n
   \r\nGenerative AI helps reduce cognitive load by summarizing, categorizing and highlighting key points, then organizing and structuring that information. In our work with clients, we have seen AI reduce the literature review timeline by 40% to 60%.
   \r\n
   \r\nFor example, "knowledge graphs" (visual representations of relationships between drugs, proteins, diseases and other entities) can help by contextualizing data, aiding in hypothesis generation and decision-making, as well as predicting relationships between entities and drug repurposing during target prioritization.
7. \r\n

\r\n

AI integration advancements lead to increased opportunities

\r\n

In the past, integrating AI into the development process was challenging, as the tools were complex and required advanced knowledge of coding. Today's AI tools are automated and easy to use, featuring pre-trained, pre-configured models, frameworks and drag-and-drop AI pipeline-building platforms. These make AI accessible to everyone involved in the drug discovery process, including medicinal chemists and pharmaceutical scientists.

\r\n

Although AI is still relatively nascent, integrating it into existing drug discovery initiatives will rapidly enhance the pipeline and unlock new opportunities in the market for organizations that leverage it. Keeping pace with the leading industry trend of democratizing AI and leveraging an AI-driven integrated discovery and analysis platform can effectively optimize the drug development process and facilitate decisions that result in improved outcomes.
\r\n 

\r\n"}}" id="text-599cb31d74" class="cmp-text">

Development of a new drug often takes 10 to 15 years, costs billions of dollars-and fails close to 90% of the time, according to one recent study. Life sciences companies must do better if they are to thrive in an era of spiraling costs, increased competition, and pressure to deliver new treatments more quickly.

Historically, drug development followed a linear path. Researchers would identify a target, screen existing compounds and optimize them to create new drugs. Artificial intelligence-in which advanced algorithms analyze vast data sets to suggest new solutions to problems-can collapse that linear path, blurring the sequential lines between discovery and analysis.

Today's AI-driven drug discovery platforms do this by integrating various public, commercial and custom catalogs for virtual screening. Leveraging analysis and visualization capabilities and advanced features like generative AI-based models, these platforms effectively expedite the generation of novel molecules. The most important benefit, however, might just be how these AI platforms expand the accessibility of that data-extending it across all stakeholders for seamless collaboration.

Small wonder, then, that one report found 84% of current users and 70% of current non-users expect AI to drive significant or even transformative impact in drug discovery over the next five years. Here are some of the areas in which that impact is (or will be) most apparent:

1. Automated discovery platforms
   
   The drug discovery process has evolved from manual methods to advanced automated techniques. Think of an AI-powered drug discovery platform as a "prime mover," or source of propulsion, that links and unleashes the combined power of the drug discovery ecosystem-providing insights, predictions and solutions that were previously challenging to obtain.
   
   Modern computational strategies have significantly accelerated discovery processes. Target structure prediction, for example, anticipates the 3-D structure of proteins. Binding site analysis uses virtual fragment simulations to identify and characterize these crucial sites. In QSAR predictions, a chemical's physical structure is used to predict its biological activity or toxicity.
   
   In the field of AI in drug development, these improvements have enabled precise prediction of pertinent physiochemical processes and ADMET (absorption, distribution, metabolism, excretion and toxicity) properties, which can reduce the drug failure rate due to toxicity and safety issues. In as little as 40 days, researchers can now identify, synthesize and test 20 candidates with a 10% success rate using AI, compared to 1% in the past.
   
   
2. AI for predicting protein structure
   
   Other uses of AI include protein structure prediction using deep learning, which enables accurate 3D protein structures that aid drug design, and generating novel molecular structures through advanced AI techniques.
   
   Protein language models (PLMs) analyze protein sequences just as natural language processing (NLP) models process human language. They excel at tasks such as de novo protein sequence generation, controllable protein design, and protein property prediction.
   
   Like NLP models, PLMs are trained on vast datasets based on protein sequences that contain the instructions for creating proteins. The models learn from these sequences and gain insights into the underlying syntax of proteins to predict how different variations impact their function, stability and interactions. This enables both the virtual screening of compound libraries for drug candidates and the assessment of potential side effects on candidates based on their protein interactions.
   
   Sparked by the commercial success of RNA vaccines for COVID-19, biotechnology firms are now pursuing therapeutics based on engineered circular RNA (circRNA). RNA, in its usual linear form, is short-lived. However, circRNA's increased stability could enhance its therapeutic potential, even at low-dose levels. AI-based methods such as DRfold have improved the accuracy of RNA models by more than 70%, which can aid in the design of RNA-targeted small molecules. It must be noted that while DRfold has great potential, it remains somewhat unproven.
   
   RNA molecules can adopt specific 3D motifs that are considered druggable, offering untapped potential to therapeutically modulate numerous cellular processes, including those linked to protein targets previously considered "undruggable."
   
   
3. Literature search and knowledge management hubs
   
   The influx of information during the literature review portion of drug discovery research can be overwhelming. AI-based tools can help by leveraging technologies such as text mining and NLP to automate key pieces.
   
   Generative AI helps reduce cognitive load by summarizing, categorizing and highlighting key points, then organizing and structuring that information. In our work with clients, we have seen AI reduce the literature review timeline by 40% to 60%.
   
   For example, "knowledge graphs" (visual representations of relationships between drugs, proteins, diseases and other entities) can help by contextualizing data, aiding in hypothesis generation and decision-making, as well as predicting relationships between entities and drug repurposing during target prioritization.

AI integration advancements lead to increased opportunities

In the past, integrating AI into the development process was challenging, as the tools were complex and required advanced knowledge of coding. Today's AI tools are automated and easy to use, featuring pre-trained, pre-configured models, frameworks and drag-and-drop AI pipeline-building platforms. These make AI accessible to everyone involved in the drug discovery process, including medicinal chemists and pharmaceutical scientists.

Although AI is still relatively nascent, integrating it into existing drug discovery initiatives will rapidly enhance the pipeline and unlock new opportunities in the market for organizations that leverage it. Keeping pace with the leading industry trend of democratizing AI and leveraging an AI-driven integrated discovery and analysis platform can effectively optimize the drug development process and facilitate decisions that result in improved outcomes.

Latest posts