With its potential for significant cost and time reductions, higher regulatory success rates, and accelerated market entry, drug repurposing is emerging as a compelling strategy in pharmaceutical development. As artificial intelligence (AI) continues to weave its way into reprofiling frameworks, these advantages are poised to expand further, promising more streamlined processes and increased discovery efficiencies. Amidst these high expectations, the critical question remains: What does it take for AI to live up to its potential in transforming drug repurposing efforts?

This article dives into critical considerations that can make or break the success of your AI-assisted drug repurposing projects.

Dark data and accessibility: Is your data ready for AI?

It’s not a secret: AI analytics draws its power from data. Fueled by massive volumes of information, AI can uncover trends and patterns often missed by human analysts and generate actionable insights to accelerate your reprofiling projects.  

Could your most insightful data be out of reach?

Accounting for an estimated 55% of all organization knowledge, “dark data” refers to unstructured information that companies accumulate over time but never use for other purposes. Typically, unharmonized and stored within disjointed systems, this data has limited accessibility, which impedes its integration into operational efficiencies and strategic AI analytics that can move repurposing efforts forward.

Dark data can include:

• Unpublished and archived clinical trial reports like abandoned trials.  

• Real-world data (RWD) like insurance claims, electronic health records (EHRs), and patient histories.

As recent usages demonstrate, dark data combined with AI technologies has the potential to successfully uncover new therapeutic uses for existing drugs. However, organizations failing to leverage dormant data to support their reprofiling endeavors can transform a promising strategy into a missed opportunity.

In 2020, Insitro and Bristol Myers Squibb used an AI-supported screening platform and archived clinical trial data to identify novel targets for amyotrophic lateral sclerosis (ALS) and frontotemporal dementia.

‍

Data diversity and prediction accuracy: Can you trust your AI models?

Data diversity is crucial in AI model accuracy, particularly in high-stakes fields like drug repurposing. AI models trained and operating on limited data can affect your analytics and the quality of generated insights, undermining your strategy benefits and hindering drug repurposing workflows.

Lack of data diversity can result in:

• Poor model validation and limited abilities— Training AI models on homogeneous data restricts their learning to a narrow set of scenarios and limits pattern recognition. This can lead to overlooked critical insights and may hinder the model's ability to apply findings broadly, impeding drug repurposing efforts. ‍
• Bias and misleading insights— If fueled with biased or unrepresentative data, AI algorithms risk generating skewed insights that lead to poor decision-making. This can result in drug repurposing recommendations that are less effective or unsafe for unrepresented groups, potentially causing adverse outcomes in clinical applications. ‍
• False positives and wasted resources— Relying on AI predictions based on biased data can result in false positives, where non-viable drug reprofiling candidates are mistakenly pursued. This not only wastes time and financial resources but also diverts attention from more promising leads.

Integrating dark data into training and analytical datasets can enhance data diversity and help circumvent information bias. Yet, to strengthen predictive algorithm capabilities, organizations must extend beyond their internal resources to enrich their databases, bolster AI model robustness, and set a breeding ground for precise predictions that expedite drug repurposing efforts.

In 2024, researchers leveraged machine learning models over a decade of RWD to emulate thousands of clinical trials and have identified five drug repurposing candidates to treat Alzheimer's disease.  

‍

Data quality and reliability: How strong is your data foundation?

Having a wealth of diverse data can significantly enhance the potential of AI in pharmaceutical development and accelerate your repurposing pipeline. However, more data does not necessarily mean better data.

Merely accumulating large quantities of information to power AI-driven analytics isn’t sufficient to guarantee reliable and trustworthy outcomes. To achieve this, organizations must ensure that each piece of data is:

• Complete: Make sure all datasets are comprehensive, fill in missing values, and identify gaps that may affect the analysis. ‍
• Accurate: Verify that data reflects reality and correct any errors or anomalies found during data quality assessments. ‍
• Consistent: Harmonize formats, values, and structures across all data sources to facilitate analysis and integration. ‍
• Relevant: Align data with specific research goals or project needs so it directly contributes to the insights or decisions being considered.

While the rise of digital technologies and data-sharing initiatives allow for simplified access to massive volumes of information, you must verify and clean data for your AI-driven reprofiling strategies to ensure positive outcomes and avoid costly mistakes.

Glimpse into a real-life example: Drug discovery company Insilico Medicine collaborated with top institutions, including the Mayo Clinic and Harvard Medical School, and gained access to extensive, well-curated clinical data to enhance its AI-driven platforms. This led to new drug repurposing opportunities for diseases like ALS.

‍

Knowledge management: The backbone of successful AI-assisted drug repurposing

As drug repurposing frameworks become more reliant on cognitive technologies, organizations must reassess how they collect, clean, and store information to address data-related challenges and maximize AI's potential. That is why knowledge management strategies have grown even more critical in recent years.

By investing in the right knowledge management solution, you can ensure data availability, diversity, and quality, enabling seamless integration into AI-driven drug repurposing strategies. However, selecting and deploying the right approach that aligns with your project needs and objectives requires a deep understanding of knowledge management options to overcome data integration barriers and propel your reprofiling endeavors to new heights.

Download this white paper to explore effective knowledge management solutions

‍

Successful AI-driven drug repurposing strategy: It all comes down to data

As AI takes center stage in drug repurposing workflows, pharmaceutical companies must scrutinize the data that powers their models to avoid misguided investments and reduce related risks. By understanding key elements of advanced data-driven technologies, organizations can craft and deploy effective knowledge management solutions to strengthen their AI-driven drug repurposing strategy, laying the foundation for therapeutic breakthroughs that will transform the lives of millions of patients.