The convergence of machine learning, deep learning and genomics, especially in the area of AI-powered genomic health prediction, while remarkably promising will also present remarkably challenging unintended consequences. A recent report suggests areas which need to be explored – starting now – as “the issues posed by the…technologies become harder to predict, more complex and more numerous.”
DNA.I.: Early findings and emerging questions on the use of AI in genomics
AI and genomics futures is a joint project between the Ada Lovelace Institute and the Nuffield Council on Bioethics that investigates the ethical, and political economy issues arising from the application of AI to genomics – which [the authors] refer to throughout [the] report as “AI-powered genomics“.
From the Report
- AI-powered genomics has seen significant growth in the past decade, driven principally by advances in machine learning and deep learning, and has developed into a distinctive, specialised field.
- Private-sector investment in companies working on AI-powered genomics has been substantial – and has mainly gone to companies working on data collection, drug discovery and precision medicine.
- The most prominent current and emerging themes in research on AI-powered genomics relate to proteins and drug development, and the prediction of phenotypic traits from genomic data.
- According to P&S Intelligence, economic forecasts have suggested the market for AI and genomics technologies could reach more than £19.5 billion by 2030, up from half a billion in 2021.
The increasing convergence of AI and genomics is set to present policymakers with a new set of practical and theoretical challenges. Considered separately, developments in AI and in genomics already pose deep questions concerning agency, privacy, quality, bias and power. Considered in relation to one another, the issues posed by the two technologies become harder to predict, more complex and more numerous.
While there has been much research considering the ethical impacts of AI and genomics as separate technologies, comparatively little attention has been paid to exploring the broader implications of the two technologies when used together, and from a structural perspective. For policymakers seeking to navigate and regulate AI and genomics, this is a critical evidence gap.
What Next?
- The specific combination of emerging themes and capabilities identified in AI-powered genomics points to the increasing viability of two broad techniques within healthcare over the next five to ten years:
- AI-powered genomic health personalisation: the ability to understand how treatment for the same health condition might vary between different people on the basis of genomic variations, and to tailor and adapt treatments accordingly.
- AI-powered genomic health prediction: the use of genomic data to estimate the probability of different people developing particular health conditions, responding well or badly to particular medicines or treatments, or being affected by lifestyle factors.
- The potential emergence of these techniques raises profound, urgent ethical, legal and policy questions.
- While some of these issues are already discussed and accounted for in existing legal, ethical and policy discourse, there are many questions concerning the macro-level impacts of developments in AI-powered genomics that have yet to be adequately explored.
- In particular, there is an urgent, relatively unmet need for sustained thinking and research on the structural, political, and economic implications of AI-powered genomic health prediction, and how its development might be steered and governed in line with public values and priorities.
“What are the examples of applying AI techniques to genomic data for drug discovery?”
(Generated using OpenAI’s ChatGPT 3.5)
AI techniques, particularly machine learning, have been increasingly applied to genomic data for drug discovery.
These examples demonstrate how machine learning’s ability to handle and analyze large-scale genomic data is revolutionizing drug discovery by enabling researchers to make data-driven decisions and uncover hidden insights that might not be apparent through traditional methods.
Here are some examples of how machine learning powers biobank-driven drug discovery:
- Target Identification and Validation: Machine learning models can analyze genomic data to identify potential drug targets by identifying genes associated with specific diseases. This involves integrating genetic, expression, and clinical data to prioritize genes that play a role in disease pathways.
- Biomarker Discovery: Genomic data can be used to discover biomarkers that indicate disease presence, severity, or response to treatment. Machine learning algorithms can identify patterns in genomic data that are associated with specific clinical outcomes.
- Drug-Target Interaction Prediction: Machine learning models can predict the interactions between drug molecules and target proteins by analyzing genomic and proteomic data. This helps in identifying potential drug candidates that could modulate disease-related proteins.
- Drug Repurposing: AI can help identify existing drugs that could be repurposed for new indications. By analyzing genomic data and drug profiles, machine learning algorithms can find matches between the molecular mechanisms of drugs and disease pathways.
- Patient Stratification: Genomic data can enable the identification of patient subgroups based on genetic factors. Machine learning algorithms can help in stratifying patients to determine who is more likely to respond positively to a specific treatment.
- Virtual Screening: AI models can be used to screen large chemical compound libraries for potential drug candidates. By integrating genomic data with chemical structure data, machine learning algorithms can predict the likelihood of a compound binding to a target protein.
- Clinical Trial Design: Machine learning can assist in designing more efficient clinical trials by analyzing genomic data to identify patient populations that are more likely to respond to a new drug. This can reduce costs and accelerate the drug development process.
- Drug Toxicity Prediction: Genomic data can provide insights into potential adverse drug reactions. Machine learning models can predict the likelihood of toxic reactions based on genetic variations and other factors.
- Multi-Omics Integration: Machine learning algorithms can integrate data from various omics sources, such as genomics, transcriptomics, proteomics, and metabolomics. This holistic approach can provide a comprehensive understanding of disease mechanisms and potential drug targets.
- Data Mining Biobanks: Biobanks contain large collections of genomic and clinical data. Machine learning techniques can extract valuable insights from these data repositories, aiding in the discovery of new drug targets and biomarkers.
- Predicting Drug Response: By analyzing genomic data from patients who have undergone specific treatments, machine learning can predict individual patient responses to certain drugs, leading to more personalized treatment strategies.
“What are the Expectations for AI in healthcare?”
The expectations for AI in healthcare are quite high, with the technology being seen as having the potential to revolutionize various aspects of the healthcare industry. It’s important to note that while these expectations are promising, the successful implementation of AI in healthcare also requires addressing challenges such as data privacy, regulatory compliance, bias in algorithms, and establishing effective human-AI collaborations.
Here are some of the key expectations for AI in healthcare:
- Enhanced Diagnostics: AI is expected to improve the accuracy and speed of medical diagnosis by analyzing complex medical images (such as MRIs, CT scans, and X-rays) and identifying patterns or anomalies that might be difficult for human clinicians to detect.
- Personalized Treatment: AI can help tailor treatments to individual patients based on their unique genetic makeup, medical history, and other factors. This includes predicting how patients might respond to certain medications and interventions.
- Drug Discovery and Development: AI is anticipated to accelerate drug discovery by analyzing vast datasets and predicting potential drug candidates. It can also help optimize clinical trial design, potentially speeding up the development of new therapies.
- Administrative Efficiency: AI has the potential to streamline administrative tasks in healthcare institutions, such as scheduling appointments, managing electronic health records, and automating billing processes.
- Remote Monitoring and Telemedicine: AI-powered wearable devices and remote monitoring tools can continuously track patients’ health metrics and alert healthcare providers to potential issues. This is particularly relevant for chronic disease management and telemedicine services.
- Predictive Analytics: AI can analyze patient data to predict disease outbreaks, patient readmissions, and potential health risks, allowing healthcare systems to allocate resources more effectively.
- Genomic Medicine: AI can analyze genomic data to identify disease risk, potential treatment options, and personalized preventive measures.
- Robot-Assisted Surgery: AI-enabled robots can assist surgeons with precision and accuracy during complex procedures, potentially reducing complications and improving patient outcomes.
- Natural Language Processing (NLP): NLP techniques can help extract valuable information from medical texts, such as clinical notes and research papers, aiding in evidence-based decision-making.
- Patient Engagement and Education: AI-driven chatbots and virtual assistants can provide patients with medical information, answer queries, and offer personalized health advice.
- Ethical Decision Support: AI can assist clinicians in making complex ethical decisions by analyzing patient data and recommending treatment options based on a combination of medical evidence and ethical principles.
Additional Resources
Technology Convergence and Market Disruption: Rapid advancements in technology are changing market dynamics and user expectations. See: Disruptive and Exponential Technologies.
AI Discipline Interdependence: There are concerns about uncontrolled AI growth, with many experts calling for robust AI governance. Both positive and negative impacts of AI need assessment. See: Using AI for Competitive Advantage in Business.
Track Technology Driven Disruption: Businesses should examine technological drivers and future customer demands. A multi-disciplinary knowledge of tech domains is essential for effective foresight. See: Disruptive and Exponential Technologies.
