By Deepti Dubey, Ph.D. and Harsha K Rajasimha, Ph.D.

In the realm of clinical trials and rare disease research, the intersection of Artificial Intelligence (AI) and Machine Learning (ML) holds the promise of transforming patient recruitment, data collection, and analysis. However, the current landscape is marked by challenges, particularly the lack of diversity in patient populations and datasets. It is important to shed light on the risks posed by biased data in AI algorithms and the ongoing efforts to foster global collaboration for more inclusive clinical trials.

Challenges of Lack of Diversity in Clinical Trials:

A historical dearth of diverse genetic and morphological data presents operational challenges in clinical trials, impeding drug development. The exorbitant costs (approx. $2.5 billion per drug) and a high failure rate (nine out of 10 drugs) contribute to more prolonged development timelines for rare diseases than for common indications. The time gap between gene mutation discovery and the corresponding treatment being approved by regulators is unsustainable, prompting the need for a paradigm shift. Ethical, sociological, and economic consequences, highlighted in a recent National Academies report, project significant financial losses over the next 25 years due to reduced life expectancy and productivity among underrepresented clinical trial populations¹.

The lack of diversity in clinical trial participants contributes to gaps in understanding diseases, preventive factors, and treatment effectiveness. FDASIA Section 907 emphasizes the necessity of diverse populations in clinical research for safer and more effective medical products. However, underrepresentation persists due to financial barriers, lack of awareness, and limited access to trial sites. Ongoing challenges with trust, transparency, and consent further hinder recruitment from disadvantaged or minority groups².

AI's Potential to Accelerate Clinical Trials:

AI has emerged as a potential solution to the repetitive challenges faced by clinical researchers. By automating patient recruitment, streamlining data collection, detecting anomalies, data inconsistencies/errors, and automating data analysis, AI significantly reduces time and costs in clinical trials. Improved efficiency in patient recruitment and protocol design is expected to enhance trial success. AI-based monitoring and analysis positively impact result measurement and interpretation. These

tools can integrate diverse data types, matching patients with complex inclusion criteria, and broaden access to trial information. During trial conduct, AI-driven sensors and wearable devices improve patient monitoring, while AI tools facilitate comprehensive statistical analysis, addressing challenges like missing data and visits. AI holds great promise in advancing drug discovery and clinical trials, especially in areas with limited profitability, such as rare diseases and targeted therapies³.

AI's Role in Improving Diagnoses:

From screening newborns to navigating genetic information complexities, AI plays a crucial role in disease diagnosis. In rare diseases, accurate diagnoses are pivotal for patient triage, risk assessment, and targeted therapies. Traditional diagnostic approaches involve a comprehensive assessment of medical history, physical exams, and genetic testing, a process that can take years. AI-based approaches, like machine learning

(ML) algorithms, analyze extensive datasets to identify distinctive patterns linked to specific rare diseases. Deep learning (DL) models aid diagnostic decisions through phenotypic characterization⁴. Knowledge graphs, utilizing historical data and medical knowledge, effectively classify diseases⁵. DL-based approaches, such as gait analysis in Huntington's Disease (HD), gauge disease severity⁶. For e.g., DeepMind's AlphaMissense predicts the molecular effects of genes and mutations, enhancing the identification of pathogenic mutations and unknown disease-causing genes, promising a significant boost in diagnostic yield for rare genetic diseases⁷. AI's diagnostic prowess underscores the necessity of a diverse dataset for optimal algorithm training.

The pressing need for diversity in clinical trial data, particularly in rare diseases, raises concerns about the applicability of AI models to diverse patient cohorts. Most algorithms in the rare disease domain are trained on either small datasets, predominantly comprising patients from North America and some European countries, or larger genomic datasets that lack diversity⁸.

Strategies for Improving Data Access and Collection:

In rare diseases, where the patient pool is often small, patient-centric studies are imperative. Modern and Decentralized clinical trials (DCT) have significant potential to enhance access and data collection, supporting global patient participation without repeated travel burdens. Real-world data from patient registries, natural history studies, and existing medical records offer crucial insights for drug development. Overcoming regulatory barriers to data sharing, ownership, and cross-border collaboration is crucial. Generative AI and large language models (LLMs) have the potential to overcome language and cultural barriers to engage underserved communities in clinical research. Establishing global clinical trial networks and advocating for diversity, equity, inclusion, and access in clinical trials are necessary steps. Crafting collaboration agreements and fostering partnerships at policy and project levels are essential. The intricacies of data protection, ethics committees, and sharing agreements demand concerted efforts for truly equitable cross-border collaboration.

The journey toward more inclusive clinical trials, powered by AI, requires a comprehensive approach addressing data diversity, cross-cultural collaboration, and equitable global participation. The IndoUS bridging RARE Summit exemplifies the commitment of stakeholders to foster collaboration and overcome challenges in rare disease research. Moving

forward involves not only technological advancements but also a collective effort to ensure that the benefits of AI are accessible to diverse patient populations worldwide. Bridging the gaps in clinical trials is not just a scientific imperative but a global responsibility to ensure healthcare innovations are inclusive and accessible to all.

Deepti Dubey, Ph.D., is an accomplished researcher with expertise in Molecular Genetics, Neurological Disorders, and Rare Diseases from esteemed institutions. In collaboration with non-profit patient organizations, she has driven research initiatives for accelerating breakthroughs for rare diseases. With a Clinical Trials Specialization, Deepti is a scientific writer at IndoUSrare.

Harsha K Rajasimha, Ph.D., is the Founder of Indo US Organization for Rare Diseases (https://IndoUSrare.org), a non-profit organization with the mission to build collaborative bridges between various stakeholders of rare disease research between the US and India, and Founder of Jeeva Clinical Trials (https://jeevatrials.com), a venture-backed startup with the mission to modernize clinical trials, dramatically improve efficiency and universal accessibility. Harsha chairs the annual Indo US bridging RARE Summit (https://summit.indousrare.org to bring the stakeholders together to address grand challenges.

References:

1.https://www.statnews.com/2022/08/03/lack-of-diversity-in-clinical-trials-costs-billions-of-dollars-incentives-can-spur-innovation/

2. A roadmap to increase diversity in genomic studies. Nat Med. 2022 Feb;28(2):243-250
3. Artificial Intelligence Applied to clinical trials: opportunities and challenges. Health Technol (Berl). 2023;13(2):203-213.
4. The Impact of Artificial Intelligence in the Odyssey of Rare Diseases. Biomedicines. 2023 Mar 13;11(3):887.
5. Improving rare disease classification using imperfect knowledge graph. BMC Med Inform Decis Mak. 2019 Dec 5;19(Suppl 5):238.
6. A Deep Learning-Based Approach for Gait Analysis in Huntington Disease. Stud Health Technol Inform. 2019 Aug 21;264:477-481.
7. Accurate proteome-wide missense variant effect prediction with AlphaMissense. Science. 2023 Sep 22;381(6664):eadg7492.
8. A roadmap to increase diversity in genomic studies. Nat Med. 2022 Feb;28(2):243-250