- AI Grand Rounds
- Posts
- Rapid Progression of AI in Genomics and Drug Discovery
Rapid Progression of AI in Genomics and Drug Discovery
How AI can bring newer, better and personalized drugs to us faster
Fellow Healthcare Champions,
Are you overwhelmed by all the fluff and hype around AI and not sure how to identify meaningful information? We get it. As busy clinicians ourselves, our newsletter, "AI Grand Rounds," is here to provide clinically relevant AI information.
No matter who you are—a healthcare provider, inventor, investor, or curious reader—we PROMISE one thing: you will always find information that is relevant, meaningful, and balanced.
Let’s join our journey to share clinically meaningful and relevant knowledge on healthcare AI.
Your fellow Physicians,
Table of Contents
🚨 Pulse of Innovation 🚨
Breaking news in the healthcare AI
The Impact of AI in Genomics and Drug Discovery
The conventional drug discovery process is a decade-long process, that is primed for innovation. AI plays a crucial role in genomics research by improving the analysis of complex genetic data. It uses machine learning and deep learning to identify patterns and predict genetic mutations, aiding in the understanding of genetic diseases, enhancing diagnostic accuracy, and supporting personalized treatment plans. AI-driven genomic analysis tools are increasingly being integrated into clinical settings to assist healthcare professionals in patient care decisions.
Foundational AI Use Cases in Genomics:
AI applications not only enhance the efficiency and accuracy of research but also pave the way for innovative therapeutic solutions. Here are some key areas where AI is making significant contributions:
Genomic Data Analysis
Drug Discovery and Development:
Precision Medicine
CRISPR and Genome Editing
Integration into Clinical Settings:
Recent advances include AlphaFold by DeepMind, which predicts protein structures for over 200 million proteins, aiding in drug target identification. Another is Insilico Medicine's Phase I clinical trials for the first AI-discovered molecule, showcasing AI's potential in drug development.
These use cases demonstrate AI's pivotal role in enhancing genetic research and pharmaceutical development, offering innovative solutions that promise to transform patient care and therapeutic practices
The global market for AI in Genomics is rapidly expanding, with software as the largest segment for now, but it is anticipated that the genome sequencing segment will reach $5.6 billion in 2032.
🧑🏼🔬 Bench to Bedside👨🏽🔬
Developments in healthcare AI research and innovations
Personalized Atrial Fibrillation Management: Evaluating Anti-arrhythmic Efficacy with Digital Twin Technology
Early rhythm control with antiarrhythmic drugs is recommended for atrial fibrillation (AF), yet safety concerns often outweigh efficacy in current guidelines. With advances in digital twin technology, researchers are now exploring personalized approaches to AF management. A recent single-center study published in Nature Digital used virtual models of patients’ left atria to simulate responses to amiodarone at varying doses, offering insights into drug effectiveness.
Key Findings:
Patient-specific digital twin models were created by combining CT imaging and electro-anatomical mapping data, enabling virtual AF induction and adjustment of amiodarone levels to assess AF termination.
Patients were classified into effective and ineffective groups based on virtual AF termination at therapeutic amiodarone doses.
Source: Nature Digital
The one year clinical outcomes after AF ablation showed significantly better results for patients in the effective group compared to the ineffective group, with AF recurrence rates of 20.8% vs. 45.1%.
Limitations:
Limitations of the study include a model that only represents the left atrium, overlooking interactions between the atria that influence atrial fibrillation. Additionally, selection bias and concerns about reproducibility in broader populations may impact the generalizability of findings. The study also did not consider the potential adverse effects of amiodarone, which are crucial for evaluating overall treatment efficacy and safety.
Conclusion:
Digital twin technology offers a powerful, personalized tool for evaluating anti arrhythmic drugs, potentially reducing adverse events by predicting individual drug responses.
If automated through AI, this approach could enable non-invasive assessments and broader clinical applicability.
Source: Nature Digital
🧑🏽⚕️ AI in Clinic 🏥 |
Developments in healthcare AI research and innovations |
PathAI: Revolutionizing Diagnostics Through AI |
PathAI is a Boston-based company that’s pushing the boundaries in pathology through advanced artificial intelligence solutions.
Founded in 2016 by Dr. Andy Beck, a pathologist and former Harvard Medical School faculty member, and Aditya Khosla, an MIT-trained AI researcher, PathAI’s mission is to bring precision and efficiency to pathology diagnostics.
By combining medical expertise with cutting-edge AI, PathAI has quickly positioned itself as a leader in digital pathology, catering to both clinical diagnostics and pharmaceutical research.
Key Features and Advances
PathAI’s technology analyzes high-resolution pathology slides, automating and enhancing the accuracy of cancer and other disease diagnoses. Some of its key innovations include:
Deep Learning Models for Pathology: PathAI’s algorithms are trained on millions of annotated slides, learning to detect cancerous and precancerous cells with a level of precision that reduces human error.
Platform for Diagnostic Assistance: By providing a second opinion to pathologists, PathAI ensures that diagnostic accuracy improves, benefiting both practitioners and patients.
Pharmaceutical and Clinical Research Support: PathAI has partnerships with major pharma companies to streamline drug discovery, especially in oncology, by using AI to analyze tissue samples and identify biomarkers faster than traditional methods.
Regulatory Compliance and Data Security: With HIPAA-compliant systems and rigorous data security, PathAI meets the high standards required for handling sensitive medical data.
PathAI’s recent developments include PLUTO, a foundational AI model trained on over 160,000 pathology images across various disease areas.
This model supports a broad range of tasks, from cell-level analysis to whole-slide pathology interpretation, by leveraging a diverse dataset of disease types and specimen details.
PLUTO enhances PathAI’s existing suite of tools by reducing costs and speeding up model development, providing scalable solutions for biopharma and pathology labs.
Furthermore, PathAI’s recent partnership with Quest Diagnostics has integrated digital pathology solutions, enhancing diagnostic speed and expanding AI use in large-scale pathology services.
Limitations
PathAI’s focus remains in research and development, and while its tools advance diagnostic pathology, they are not yet available for routine clinical use.
Implementing these technologies across diverse lab settings requires substantial infrastructure, and the dependence on high-quality data may limit efficacy in lower-resourced regions.
The Future
Looking forward, PathAI’s focus on high-performance foundational models like PLUTO and strategic partnerships aims to create new AI-powered tools for widespread clinical and research use.
With continued advancements, PathAI could help bridge global healthcare gaps by enabling remote diagnostics and enhancing AI integration in routine pathology, transforming the future of diagnostic medicine.
🤖 Patient First, AI Second🤖 |
Ethical and Regulatory Landscape of Healthcare AI |
Racial differences in laboratory testing as a potential mechanism for bias in AI
This study investigates racial disparities in laboratory testing within emergency departments (EDs) at two major U.S. academic hospitals, Beth Israel Deaconess Medical Center (BIDMC) and the University of Michigan (U-M).
It highlights a crucial ethical challenge in AI: the risk of amplifying biases that already harm racial minorities.
Past research shows Black patients often face disparities in receiving care, such as lower colon cancer screening rates and higher all-cause mortality.
In the context of AI, these inequities could be unintentionally embedded in models, particularly if these models rely on biased correlations or historical healthcare data without appropriate safeguards.
Given the rise of AI in healthcare, the study also explores how these disparities might affect the predictive accuracy and fairness of AI systems.
Key Findings
Testing Disparities in Unmatched Cohorts: Initial analyses without adjustments found that Black patients were significantly less likely to receive key tests, such as CBCs and metabolic panels, compared to White patients. These differences persisted across both institutions, suggesting broader systemic issues rather than hospital-specific practices.
Effects of Matching on Disparities: After adjusting for clinical factors like age, sex, and triage score, some disparities narrowed but remained significant. White patients were still more likely to receive tests such as CBCs, while Black patients had marginally higher troponin testing rates, indicating that clinical adjustments alone may not fully account for racial disparities.
Admission Rates and Testing Patterns: White patients had a higher hospital admission rate, which partly explained testing differences, as admitted patients generally undergo more extensive testing.
Ethical Issues and Future Implications
The persistence of these disparities presents ethical concerns, especially for AI applications.
AI models trained on data with racial biases risk reinforcing and perpetuating these disparities, potentially leading to systematic undertesting or misdiagnosis of Black patients. For instance, models trained on data where Black patients receive fewer diagnostic tests may “learn” that Black patients are at lower risk for certain conditions simply because they lack the test results that would indicate otherwise.
Addressing these biases requires more than omitting race from model inputs, as proxy variables can continue to introduce bias.
The study suggests exploring causally motivated approaches and counterfactual fairness techniques to ensure that AI models account for bias and do not reinforce existing disparities.
Disclaimer: This newsletter contains opinions and speculations and is based solely on public information. It should not be considered medical, business, or investment advice. This newsletter's banner and other images are created for illustrative purposes only. All brand names, logos, and trademarks are the property of their respective owners. At the time of publication of this newsletter, the author has no business relationships, affiliations, or conflicts of interest with any of the companies mentioned except as noted. ** OPINIONS ARE PERSONAL AND NOT THOSE OF ANY AFFILIATED ORGANIZATIONS!
Reply