This is the first instalment of our blog post series Reimagining Genetic Testing: Can AI Do it All?
Join us for a thought-provoking blog series as we delve into the potential and limitations of artificial intelligence in genetic testing. Discover how AI can revolutionize:
✅ Understanding the power of clinical context
✅ Simplifying interpretation and enhancing explainability
✅ Transforming reporting for both specialists and patients
Genetic testing labs often operate with one hand tied behind their backs due to limited access to clinical context, which impacts both the accuracy and speed of variant interpretation. This is because the clinical context provided in routine genetic diagnosis is often sparse, limited to a few notes or a brief paragraph. However, the other end of the spectrum is equally impactful, as lengthy reports with critical information buried in unstructured text can be just as problematic, creating barriers to effective analysis.
Imagine the possibilities if labs had access to a complete clinical picture. With richer context, different disease phenotypes could more accurately be linked to specific variants or patterns. Currently, variants are related to diseases using a transitive property: if a variant is in a gene known to be related to a disease with a generic phenotype, we assume a connection. There are some complexities here in reality, such as molecular mechanisms and other modifiers of this process. However, the process is generally applied. There are a few exceptions to this simple rule in AION to help our users to quickly solve cases, but with better clinical context, many more relationships would become statistically significant. And this is just one of the many potential improvements.
So, why is this information not readily available? Let's take a look at the roadblocks hindering progress and explore how recent developments in AI could offer a solution.
The challenges inherent in the interaction between healthcare providers (HCPs) and genetic testing laboratories create a significant bottleneck in the delivery of efficient and timely diagnostic genetic testing for rare diseases. These systems are siloed and were not designed for interconnectivity. This multifaceted problem is rooted in several key issues:
1. Strict Healthcare Workflows: Established healthcare workflows, while essential for maintaining standards and ensuring patient safety, can be rigid and resistant to change. These entrenched processes can hinder the adoption of new technologies and innovative approaches to data sharing, which are crucial for streamlining genetic testing. Changing this is very difficult.
2. Data Protection Concerns: HCPs are understandably cautious about sharing Personal Identifiable Information (PII) due to the sensitive nature of genetic data and the potential for misuse. While these concerns are valid, the lack of clear regulatory guidelines and standardized data sharing practices has further exacerbated the issue, leading to reluctance and delays in data exchange.
3. Technological Limitations: Until recently, the technological tools for anonymizing and extracting data in a secure and trustworthy manner from unstructured data have been limited. This has limited the quality of data shared or even forced manual data sharing processes, which are time-consuming, prone to errors, and increase the risk of data breaches.
All of this stifles innovation in genetic testing for rare diseases. Building a shared understanding of the big potential of bridging gaps across HCPs and laboratories is key. This is the first step in generating trust and ultimately building the infrastructure that guarantees patient safety and data protection through reliable technology. The following sentence summarized well the truth behind innovation in this space: "In healthcare, innovation moves at the speed of trust."
AI and machine learning can automate data anonymization and extraction, while standardized data formats, modern technology compliant with strict cybersecurity standards and improved regulations address data protection concerns. As technology and regulations advance, fully automated genetic testing for rare diseases becomes feasible, reducing turnaround times, improving data analysis, and leading to faster diagnoses and targeted therapies.
AI-Powered Anamnesis: Imagine a scenario where AI can intelligently guide patients through a comprehensive medical history interview, similar to a knowledgeable and empathetic healthcare provider. This AI-driven anamnesis could ensure that all relevant details are captured, potentially uncovering crucial information that might otherwise be missed.
Phenotype Detective: AI can leverage its vast knowledge base to suggest potential symptoms that a patient might be experiencing but hasn't yet reported or hasn’t been explored for. By guiding physicians and patients through a targeted exploration of their symptoms, HCPs can effectively implement reverse phenotyping, leading to a more accurate variant interpretation and a more timely diagnosis.
Decoding Clinical Notes: Clinical notes often contain a wealth of unstructured data. AI can efficiently parse these notes, identify pertinent phenotypic features, and map them to standardized ontologies, such as Human Phenotype Ontology (HPO) or SNOMED terms. Current HPO taggers have issues, for instance often tagging HPO terms even when they are referring to other family members. This streamlines data analysis and facilitates interoperability between different healthcare systems. The input of tertiary analysis pipelines would become genetic data and extensive clinical notes.
Unlocking Nuanced Genetic Insights: AI can enhance genetic testing by providing richer clinical context, leading to more accurate and efficient analyses. This improved context includes factors like low penetrance (i.e., a pathogenic variant doesn't always cause the associated disease) and variable expressivity, and longitudinal phenotyping, which offer more nuanced interpretations of genetic and clinical data. Specifically, in cases of low penetrance better context accumulated of time can provide data-driven guidance into how to interpret the observed variants. Also, AI could incorporate environmental and polygenic predictions. Additionally, having phenotypic longitudinal data available would provide the same benefit as there are well known early symptomatic patterns of specific diseases that, disregarding timing may seem phenotypically very similar to other diseases. Accumulating this data over time would help interpreters give the right answer.
AI's transformative potential in rare disease genetic testing extends far beyond mere variant interpretation; it promises to redefine the entire genetic testing workflow. By helping HCPs and patients in key processes, automating others, addressing data sharing challenges, and offering deeper clinical insights, AI can significantly accelerate diagnoses, enhance accuracy, and ultimately pave the way for personalized therapies. Ultimately, AI's integration into the diagnostic odyssey will not only improve and extend the lives of patients but also revolutionize the way we approach and manage rare diseases. For variant analysts, the promise is clear: richer information creates a crisper picture of the needle variant scientists need to find in the haystack.
With all this promise, can we afford not to Reimagine Genetic Testing?
Contact us!
*Nostos Genomics regularly produces webinars, white papers, and other types of content that you may find valuable.
You can unsubscribe at any time. For more information view our Privacy Policy.