The N-of-1
Personalized genetic treatments can help Lydia and many others
The Problem - The Long Tail
There are six billion characters in our DNA. All of us have spontaneous typos in this code — a mutation. Some typos, like Lydia’s, cause serious diseases. Collectively, there are seven million people in the U.S. suffering from typos that affect the brain. Majority of affected are children. This doesn’t include the millions who have already died from these. Pre-natal genetic testing does not look for these (but they should, more on that later!). Because a typo can happen in one of billions of characters, there are only a handful of patients with the exact same one. For Lydia’s, there are only two others in the entire world. At position 683 of the gene KCNQ2, an A was mistyped as a G. That’s all it took.
This is a classic long tail problem — no mutation is common enough, but collectively there are tons. The existing Pharma approach to treat these is broken — they look for common typos and fix them with long drawn out trials. This barely makes a dent. Worse, we have put each in its own bucket and labeled them as rare, so the majority of the world feels they‘re not important to fix. How can these be rare when collectively there are millions with these mutations? The rare label is wrong and limits progress. These are not rare. These are genetic and have the same root cause. We need a systematic, platform-driven approach to fix these typos.
The Solution - An N-of-1 Platform
Monogenetic diseases are unique from the vast majority of cancers and cardiovascular diseases because we can pinpoint the exact cause — this makes them extremely treatable. There’s a groundbreaking technology called Antisense Oligonucleotides (ASOs) that can silence mutations at the source. Since the first approved drug in 2016, ASOs have successfully halted severe diseases like Spinal Muscular Atrophy and Batten’s Disease. Unlike traditional drugs that target proteins (the hardware), ASOs target at the RNA level (the code). These have repeatable properties, making customization relatively easy. By modifying a few characters of code, without changing the chemistry or dosing, you can design an ASO to target any mutation. We live at a unique moment in time where a child can be treated with a personalized ASO — an “N of 1” treatment. In 2018, Dr. Timothy Yu at Boston Children’s Hospital created the first N-of-1 ASO, skipping the lengthy clinical trial process.
Pharma is currently not set up to make money from these individual treatments — there’s no blockbuster drug to advertise, there’s no IP to protect. We started Lydian Accelerator as a non-profit to fill this gap and help accelerate this research for Lydia and others. As computer scientists, we believe in open platforms. By openly sharing the processes, tools and data from the first few N-of-1s, we want to enable more people to create these treatments. With more treatments and more data, we may be able to reduce or eliminate lab work, reducing the time and costs for each subsequent treatment. This same model could be applied as future gene editing technologies are de-risked scientifically. We’re very early but our mission is to bring the end to long tail genetic disease.
PATH to an N-of-1 ASO
Here are the sequence of steps required:
| Step | Time |
|---|---|
| 1. Genetic Sequencing. Identify the disease causing mutations in a patient's gene, as well as nearby mutations that could make good targets for genetic treatments. This is usually done as part of diagnosis. | 4 weeks |
| 2. Functional study of mutation. Depending on the type of mutation, there are several different ASO mechanisms of action. For KCNQ2, our ASO strategy is to silence an entire copy of the gene, so the patient is left with one good copy, which is enough to restore function. If no ASO mechanism is suitable, a functional study can also inform whether small molecules or gene therapies can be useful. | 4 weeks |
| 3. Start making patient derived cells. Patient derived cells are the best way to test the drugs without human trials. If you making iPSC cells, this step takes long and requires multiple sub-steps, but can be done relatively cheaply at a hire-for-contract lab. | 4-6 months (in parallel with next steps) |
| 4. Design and synthesize ASOs. ASOs have such well understood chemistry, designing and synthesizing them for screening can be quick. You can design tens or hundreds of ASO candidates to target a mutation and nearby sites. This is an iterative step, so you may want to go through multiple rounds! | 3-4 weeks ([n] times) |
| 5. Screen in patient derived cells. Screen the best candidates in patient derived cells. You look for RNA expression. If needed, you can also look for functional improvement. (for example, electrophysiology in iPSC derived neurons) | 4-8 weeks ([n] times) |
| 6. Toxicology in animals. The FDA requires abbreviated tox studies in animals before administration. This step can be done at a hire-for-contract lab but is the most expensive step. | 12 weeks |
| 7. GMP Manufacturing. Manufacture the best candidates for human administration. This step can be done at a hire-for-contract lab and we can manufacture a lifetime supply. | 8 weeks |
| 8. FDA Approval. Apply for single patient approval. The FDA is extremely forward thinking in these approaches and a strong collaborator. | 4 weeks |
| 9. Start the treatment. Dosing is scaled over a period of several months. The drug is delivered via injection to the spinal column to reach enough cells in the brain. Once we reach a chronic dose, we may need to administer the drug every 2-3 months. | N/A |
FAQ
Q: How is this different from Gene Editing or CRISPR?
We will not do justice to this in a few lines, but the short answer is that ASOs are just much further along and well understood than gene editing. They both fix the underlying cause of the disease instead of the symptoms.
ASOs have been administered in humans for several years now, gene editing is still 5-10 years away.
With ASOs, there are fewer off-target effects because of well understood chemistry and delivery mechanisms. Gene editing use viral vectors which still have challenges, especially for neurological diseases where you need to reach a large number of cells in the brain.
The promise of gene editing is that it may be a one-time treatment while ASOs need to be delivered every few months. The downside is that the burden of proof increases tremendously.
There is another alternative called Gene Replacement Therapy that has similar challenges, although is in clinic for other diseases.
Q: What are the risks? What could go wrong?
Here are a few things that could go wrong:
We could fail to identify a good ASO that targets the mutation in the manner we’d like. We are mitigating this by identifying several different locations to target the ASO (this is why we start with 100s of ASOs)
With any first-in-human trial, there are risks. We are mitigating this by using chemistries and backbones that have already been used in humans successfully.
It may be too late to restore function. This is difficult because we don’t have enough data on when these treatments stop being effective. From Spinal Muscular Atrophy, we know that the treatment was helpful at every age, but with diminishing returns as the child got older. Our hope is that future kids are able to get this kind of treatment within months if not weeks of birth.
Q: How much does this cost?
This research is astonishingly expensive today - in the single digit millions of dollars. The process is still new and not streamlined yet and the burden of proof of safety is high. Imagine doing the first few heart surgeries. The bulk of costs today are expensive animal studies to prove that each N-of-1 ASO is safe. This cost will go down substantially for subsequent treatments as we generate more data about repeatable predictable properties. There is a world where each treatment can cost a few hundred thousand dollars, that is advantageous to be covered by insurance since it will save a lifetime of higher costs.
Q: How does the N-of-1 scale?
There are three challenges of scaling this model. Scientific, Regulatory and Business model. For now, we are focused on the first one. Our hope is to prove that this is effective and safe. With more safety data and more treatments, the regulatory model will automatically scale. The business model seems like the most challenging problem right now. We need to create a streamlined system that can massively reduce the cost, timing and manual work required for each of these treatments. For a company to invest in this kind of work, they would need to create something defensible so they can extract the maximum economic benefit from their research. There is quite a lot of uncertainty here. Our goal with this non-profit model is to help play a pivotal role in the absence of significant for-profit investments in this area.