PAID CONTENT brought to you by
BACK TO HOME
How large language models are transforming pharmaceutical frontiers
Avenga CEO Yuriy Adamchuk discusses how pharmaceutical organizations can apply large language models successfully and what they need to focus on.
BACK TO HOME
PAID CONTENT brought to you by
Large language models (LLMs) are a fairly new technology, but they’ve already made their way into real-world applications in multiple industries. Despite being under huge regulatory pressure in the use of artificial intelligence (AI), pharmaceutical companies are still looking for ways to capitalize on this transformative type of AI model. This article will shed light on how pharmaceuticals are already using LLMs to their advantage and will provide tips on ensuring the success of generative AI projects.
LLMs in pharma: Accelerated drug discovery
In drug discovery, LLMs are mainly used for data analysis. This data could be scientific literature, protein sequences, and so on. For example, by obscuring parts of the protein sequence, the model can infer the missing structure, which enables the generation of new sequences.
Insilico Medicine is a company that has embraced deep learning far earlier than many of its peers. It stands out as a prominent example of effective LLM usage for drug discovery. The organization leverages different AI models, including generative AI, to enhance and expedite various drug discovery phases. One drug candidate identified through its AI platform is already in phase two of its clinical trials. Additionally, Insilico made news by employing the BioGPT LLM to discover a novel dual-target inhibitor for aging and disease.
A U.K.-based biotech company, e-therapeutics, is also worth mentioning here. It uses AI to identify new hepatocyte-associated therapeutic targets and develop RNA Interference (RNAi) medicines. In this system, LLMs find non-obvious insights across disparate datasets and can thus help accelerate the drug discovery workflow.
LLMs can also play a role in advancing research into rare diseases. Google’s DeepMind recently developed AlphaMissense, a model that analyzes DNA mutations, in a similar way to how OpenAI’s ChatGPT examines text. It was trained on a large dataset of genetic sequences, so it “knows” what normal patterns of amino acids and proteins look like and can accurately predict the impact of various mutations.
As far as business benefits go, LLMs have the potential to significantly cut drug discovery costs, accelerate time-to-market, and free up researchers to focus on more strategic tasks that require creativity.
This means the models can aid pharmaceutical organizations in boosting profitability and building a competitive edge—for example, by facilitating the development of novel therapeutic approaches.
Improved clinical trial efficiency
LLMs also hold lots of promise in clinical trials. In particular, they can assist pharma companies in designing clinical trial protocols and identifying flaws in existing ones. The models can also identify suitable candidates for clinical trials (by analyzing patient data), automatically generate clinical trial reports, and empower virtual assistant platforms to promptly answer researchers’ questions about trial protocols, data, and regulations.
Syneos, an organization that uses a suite of AI tools in its clinical trial programs, has recently inked a deal with Microsoft to leverage OpenAI developments. Syneos already leverages AI algorithms for clinical data processing and analysis. And the capabilities these LLM models bring can help fast-track timelines, optimize resource allocation, and unlock clinical trial efficiencies.
Overall, there are many advantages to integrating language models into clinical trial workflows. They’re capable of shortening trial durations, helping achieve substantial cuts in personnel and resource utilization costs, streamlining workflows, and substantially minimizing the risk of errors.
Enhanced customer engagement
One use case that hasn’t yet been covered widely but is likely to gain a lot of traction in 2024: LLMs being used for customer engagement in pharma.
Language models can promptly respond to medical inquiries. They can elevate existing virtual assistant platforms, which Pfizer, Novartis, GSK, and many others are already using, enabling them to provide up-to-date information to both patients and health care professionals faster and with a lot more accuracy.
Furthermore, pharma companies can leverage the language model’s properties to produce interactive and educational content, tailor health information to individual needs, determine the sentiment of customer feedback on social media, and engage potential clients across various platforms.
They are pretrained on a huge chunk of the web, which gives them a robust knowledge foundation for a range of pharmaceutical tasks.
They are remarkable at generalization and adaptation, so they can “learn” contexts and master new tasks quickly, such as through few-shot learning.
Interactive communication serves as another key feature as it facilitates a dynamic information exchange and integration of valuable insights, from prior knowledge or domain experts, into the models.
•
•
•
These characteristics strongly suggest that pharmaceutical organizations will increasingly adopt LLM technology in 2024.
The potential business benefits of this application are far-reaching. The LLMs’ ability to analyze previous client interactions so as to tailor content and product recommendations, and to provide health care professionals with instant access to relevant scientific literature, clinical trial data, and regulatory updates, makes them effective tools for marketing and strengthening a brand’s image. The models’ insights can positively impact companies’ ability to attract clients, provide differentiating customer experiences, and ultimately drive sales growth.
Practical tips for LLM success in pharma
1. Start with a small number of high-impact use cases
Organizations need a few high-impact use cases to verify the value of leveraging LLMs. Evaluate each of them thoroughly based upon potential value. During the assessment, the costs, complexity, and process improvements the technology can unlock should all be taken into account. Also, both technology and business functions must be involved in this exploration.
2. Establish robust governance policies
Unfortunately, even the most advanced language models occasionally slip up. They might produce inaccurate, incomplete, or at times entirely false information. Detailed governance policies protect pharma companies against the risks of using LLMs.
Each firm, whatever model it uses, must have comprehensive rules covering AI and generative AI implementation, and a framework detailing the types of data that can be input into the models. It’s best if the data is also split into categories based on sensitivity and the severity of risks associated with false predictions. Finally, the ways of leveraging the insights derived from LLM’s outputs must also be strictly regulated.
3. Work on the overall AI capabilities
To squeeze the most value possible out of LLMs, there are certain capabilities a company must establish first, as is the case for any AI project.
First, there’s no way a firm will obtain high-impact business outcomes from using the models if it can’t properly integrate and consolidate the data (trial data, biological data, patient records, scientific literature, and so on) into a single data platform. To quickly carry out testing, it should also have a resilient analytics platform.
Finally, if the firm lacks infrastructure and workflows in data collection, flow, and cleaning, it’s probably too early for it to begin experimenting with generative AI.
Final thoughts
There are three properties that make LLMs extremely promising for the pharma industry:
BACK TO HOME
This article was created by Avenga, a client of Business Reporter.
To learn more about the unique properties and benefits of AI models for pharma, click here.
Large language models (LLMs) are a fairly new technology, but they’ve already made their way into real-world applications in multiple industries. Despite being under huge regulatory pressure in the use of artificial intelligence (AI), pharmaceutical companies are still looking for ways to capitalize on this transformative type of AI model. This article will shed light on how pharmaceuticals are already using LLMs to their advantage and will provide tips on ensuring the success of generative AI projects.
LLMs in pharma: Accelerated drug discovery
In drug discovery, LLMs are mainly used for data analysis. This data could be scientific literature, protein sequences, and so on. For example, by obscuring parts of the protein sequence, the model can infer the missing structure, which enables the generation of new sequences.
Insilico Medicine is a company that has embraced deep learning far earlier than many of its peers. It stands out as a prominent example of effective LLM usage for drug discovery. The organization leverages different AI models, including generative AI, to enhance and expedite various drug discovery phases. One drug candidate identified through its AI platform is already in phase two of its clinical trials. Additionally, Insilico made news by employing the BioGPT LLM to discover a novel dual-target inhibitor for aging and disease.
A U.K.-based biotech company, e-therapeutics, is also worth mentioning here. It uses AI to identify new hepatocyte-associated therapeutic targets and develop RNA Interference (RNAi) medicines. In this system, LLMs find non-obvious insights across disparate datasets and can thus help accelerate the drug discovery workflow.
LLMs can also play a role in advancing research into rare diseases. Google’s DeepMind recently developed AlphaMissense, a model that analyzes DNA mutations, in a similar way to how OpenAI’s ChatGPT examines text. It was trained on a large dataset of genetic sequences, so it “knows” what normal patterns of amino acids and proteins look like and can accurately predict the impact of various mutations.
As far as business benefits go, LLMs have the potential to significantly cut drug discovery costs, accelerate time-to-market, and free up researchers to focus on more strategic tasks that require creativity.
This means the models can aid pharmaceutical organizations in boosting profitability and building a competitive edge—for example, by facilitating the development of novel therapeutic approaches.
Improved clinical trial efficiency
LLMs also hold lots of promise in clinical trials. In particular, they can assist pharma companies in designing clinical trial protocols and identifying flaws in existing ones. The models can also identify suitable candidates for clinical trials (by analyzing patient data), automatically generate clinical trial reports, and empower virtual assistant platforms to promptly answer researchers’ questions about trial protocols, data, and regulations.
Syneos, an organization that uses a suite of AI tools in its clinical trial programs, has recently inked a deal with Microsoft to leverage OpenAI developments. Syneos already leverages AI algorithms for clinical data processing and analysis. And the capabilities these LLM models bring can help fast-track timelines, optimize resource allocation, and unlock clinical trial efficiencies.
Overall, there are many advantages to integrating language models into clinical trial workflows. They’re capable of shortening trial durations, helping achieve substantial cuts in personnel and resource utilization costs, streamlining workflows, and substantially minimizing the risk of errors.
Enhanced customer engagement
One use case that hasn’t yet been covered widely but is likely to gain a lot of traction in 2024: LLMs being used for customer engagement in pharma.
Language models can promptly respond to medical inquiries. They can elevate existing virtual assistant platforms, which Pfizer, Novartis, GSK, and many others are already using, enabling them to provide up-to-date information to both patients and health care professionals faster and with a lot more accuracy.
Furthermore, pharma companies can leverage the language model’s properties to produce interactive and educational content, tailor health information to individual needs, determine the sentiment of customer feedback on social media, and engage potential clients across various platforms.
