Precision medicine for rare diseases AMA with Dr. Lukas Lange

‍

Host: Alexander Beadle, Science Writer and Editor at Technology Networks

Guest: Dr. Lukas Lange, CEO and Co-founder, Probably Genetic

Introduction: Lukas, our CEO and Co-founder, recently sat down with Technology Networks for an “Ask Me Anything” (AMA) addressing how artificial intelligence is transforming precision medicine for rare conditions. We're significantly reducing the time to diagnose extremely rare conditions at Probably Genetic, and time-to-diagnosis has been a major obstacle in the identification and treatment of rare diseases. We are providing an edited version of the Q&A transcript here for anyone who was not able to attend the AMA or who is seeking answers to some of the questions that continue to arise about how technology can change the landscape for patients with rare disease and those developing and launching treatments. 

‍

Q: Dr. Lange, can you share a bit about your background and what led you to found Probably Genetic?

A: Sure. I’m originally from Germany and trained as a chemical engineer before moving to the UK, where I completed a master’s in a biotech-adjacent field, and I hold a  PhD in genetics and bioinformatics from The University of Oxford. One of the most impactful projects I worked on was the 100,000 Genomes Project, which sequenced 100,000 individuals with suspected genetic diseases. That experience taught me how difficult it is for most people to even get tested. It’s only a lucky few who meet a doctor who recognizes their condition might be genetic. The majority don’t, and they endure what’s known as the “diagnostic odyssey,” which can take over seven years. Probably Genetic was founded to change that. Our goal is to help the estimated 200 million undiagnosed patients around the world get answers.

‍

Q: How does Probably Genetic’s testing program differ from traditional genetic testing?

A: Traditional testing typically starts with a physician ordering a test. But many patients never meet such a physician. Our model allows patients to find us online, often through searching their symptoms, and then complete a symptom assessment. We use an AI platform to evaluate whether their symptoms suggest a genetic disease. If the AI indicates potential, their case is referred to a telemedicine physician who decides whether a test is appropriate. If approved, the patient receives a kit, submits a sample from home, and then receives results along with a free genetic counseling session. Importantly, our service is completely free to the patient; no credit cards, no insurance. We’re funded by drug developers who are seeking to identify patients with specific conditions for treatment or clinical trials.

‍

Q: What causes the “diagnostic odyssey” for rare disease patients, and how can technology help shorten it?

A: Several factors contribute to the diagnostic odyssey for those with rare disease:

1. Slow onset of symptoms – Many children seem healthy at birth. The first signs might be developmental delays that are initially dismissed.
   
   
2. Access to physicians – Parents can’t order tests on their own. They rely on doctors who may or may not be informed or proactive.
   
   
3. Financial barriers – Genetic tests are expensive and often not covered by insurance or healthcare systems.
   
   
4. Incomplete testing – Even when tests are done, they may be limited to a narrow set of genes, missing the real cause.
   
   

Technology helps by automating and democratizing access. Technology means we can use AI to identify likely cases, telemedicine to reach patients directly, and whole genome sequencing for comprehensive analysis.

Editor’s note: A “diagnostic odyssey” is defined as the long and challenging time between the initial onset (or feeling) of symptoms and when the final diagnosis is made. The average time it takes for a patient with a rare or complex genetic disease to receive their final diagnosis is 5-7 years. 

‍

As noted in our article “10 facts about genetic testing for rare disease,” the diagnostic odyssey also includes, on average, 8 physician visits and 2 to 3 misdiagnoses, which often include unnecessary diagnostic testing or treatments. 

‍

Q: Can you explain the different types of genetic tests?

A: Absolutely. There are three main categories of genetic tests:

1. Chromosomal microarrays – Look at large, structural changes in the genome; effective for detecting things like Down syndrome.
   
   
2. Gene panels – Focused on a subset of genes relevant to a specific condition (e.g., muscular dystrophies). The downside is that they might miss the correct gene.
   
   
3. Whole genome sequencing (WGS) – Sequences the entire genome and is the most comprehensive method available today.
   
   

Editor’s note: A question was later asked about whole-exome sequencing, which Dr. Lange noted sits between panel testing and WGS, covering the protein-coding regions (about 2% of the genome).

‍

Q: How do individual genetic profiles influence treatment?

A: They can have a massive impact. Take epilepsy—generic drugs can reduce seizures, but don’t treat the cause. A specific genetic diagnosis might reveal eligibility for a gene therapy or targeted treatment that addresses the root cause, not just symptoms.

‍

Q: What are the biggest challenges in identifying undiagnosed patients?

A:

1. Awareness – Many patients don’t even know that a genetic disease is a possibility.
   
   
2. Identification – From the many who find us, only some are good candidates for testing.
   
   
3. Funding – Tests are expensive, so we built a model funded by drug developers.
   
   

Q: How is Probably Genetic using AI and large language models?

A: In two key ways:

1. Data collection – AI helps interpret patients’ natural descriptions (e.g., “my husband used to chop wood but now forgets how”) into clinical terms like “loss of executive function.”
   
   
2. Prediction – LLMs and other AI models help predict which symptoms might correspond to specific diseases. This will likely become even more effective in the future.
   
   

Q: You mentioned facial recognition—how is that used in diagnosing genetic diseases?

A: About 40–45% of genetic diseases have distinct facial features. We work with research groups that train algorithms to recognize these traits from patient photos, just as a human might recognize Down syndrome based on appearance. It’s incredibly powerful for identifying rare diseases.

‍

Q: What AI advancements are you excited about?

A:

• For diagnosis: Embedding AI into electronic health records to help doctors spot genetic red flags.
  
  
• For treatment: Using foundational models (similar to ChatGPT) to design new drugs, even generating novel protein sequences with desired properties.
  
  

Q: Can LLMs help address the issue of predefined targets in genetic testing?

A: Whole genome sequencing doesn’t rely on predefined targets—it's hypothesis-free. That said, LLMs can still help interpret unusual sequences and spot likely disease-causing variants, even among millions of differences.

‍

Q: What about disparities in access to testing?

A: There are significant disparities:

• Geographic – Patients far from major hospitals often lack access.
  
  
• Social – Women, minorities, and underserved communities face barriers.
  
  
• Economic – Tests are expensive; many can’t afford them.

Our approach with free, online access to testing is designed to remove as many of these barriers as possible.

‍

Q: How can we ensure equitable access to precision medicine?

A: If we can reduce the cost of finding patients through AI and scale that globally, it becomes cheaper to reach and treat more people. Already, we’ve had submissions from over 160 countries, proof that there is demand and a path forward for expanding access. 

‍

Q: What role does genetic counseling play?

A: Genetic counseling is essential both before and after testing. Counselors help patients understand complex results, navigate implications for family members, and prepare for difficult news. For certain conditions, like dementia, pre-test counseling is even more important due to the emotional weight of the results.

‍

Q: How have sequencing advances changed rare disease diagnosis?

A: Better sequencing = more diagnoses. Technologies like long-read sequencing are now helping in areas where short-read methods fall short, like cancer or complex structural variants. More diagnoses also mean more targeted treatments over time.

Editor’s note: Lukas also referenced a chart in pediatric seizures that best represents how sequencing can improve our understanding of variants and equate to more accurate diagnoses. That chart can be accessed here: https://www.ncbi.nlm.nih.gov/core/lw/2.0/html/tileshop_pmc/tileshop_pmc_inline.html?title=Click%20on%20image%20to%20zoom&p=PMC3&id=10088099_nihms-1869645-f0001.jpg 

‍

Q: Are certain populations underrepresented in genetic research?

A: Yes. Studies skew toward white, educated, affluent individuals. Our approach is helping fix that by reaching more representative populations globally. 

‍

Q: What are the challenges in implementing precision medicine for rare diseases?

A: It’s a long chain:

1. Identify the disease
   
   
2. Understand the biology
   
   
3. Prove it in models
   
   
4. Run clinical trials
   
   
5. Solve delivery

Each step requires time, money, and scientific breakthroughs. Add to that: small patient populations and complex regulatory paths.

‍

Q: Are there barriers to using machine learning for patient identification?

A: Yes, there are technical, financial, and behavioral barriers to using machine learning for patient identification. Integrating with hospital EHRs is tough due to data variability and privacy. Convincing physicians to change workflows is harder. And even direct-to-patient models require massive investment in marketing and trust-building. 

‍

Q: How do we bridge the gap between academic research and implementation?

A: By building data ecosystems like the UK Biobank and enabling researchers to study real-world patients. Our goal is to create that kind of platform at Probably Genetic, connecting researchers with the data and insights they need.

‍

Q: What does the future of rare disease diagnosis look like?

A: If you're sick, you find out what you have and you get treated. That’s the future we’re building toward. I think all the puzzle pieces we need exist: genome sequencing, AI, and the internet. We should be able to develop great technology that diagnoses people and get it to everyone who needs to be diagnosed.