Discover more from Superfluid
Setting the Scene for Genetic Breakthroughs
Exploring the world of gene therapy and it's challenges
Welcome to all the new subscribers that have joined us over the last few weeks! If you haven’t already, you can read my last article on why innovation is meant to be slow here.
Stay up to date with Superfluid by subscribing here:
In primary school every year we’d do Jeans for Genes day. Donate $2 and you got to wear jeans to school that day (FYI this was a big deal back then). I knew the money was going to charity, but I didn’t really know much else.
In high school, we had a small module on all things DNA and RNA, lightly touching upon genetic diseases. In all honesty, whilst I did alright in high school bio, it definitely wasn’t my favourite subject.
More recently, however, it’s all I can think about. In my last BioTech article, we explored the world of drug discovery, and the level of funding required to actually bring a drug to market.
Pulling on this thread further, unlocking the power of gene therapy has been a long-standing goal in modern medicine, offering the potential to treat and even cure genetic diseases. However, the gene therapy value chain is complex, expensive, and requires significant expertise in multiple areas of biology and manufacturing.
In this article, we'll dive into the gene therapy value chain, from target identification to clinical trials, and explore the challenges and opportunities in this field.
Let’s dive in!
The Gene Therapy Value chain
Before we go too deep, here’s a quick biology lesson about cells and genes.
Anything that is living and breathing is made up of cells. They are the smallest unit of life. Our bodies are made up of over 100 trillion cells and each of these cells is conditioned to do different functions.
We all know that our DNA is what makes us unique. Hidden inside each cell is a strand of DNA. Genes are simply a section of DNA made up of small chemicals called bases (denoted by A, C, T and G). There are about 20,000 genes inside each cell and are effectively the ‘instruction manual’ that makes your body work.
There is a standard structure for the bases within the DNA, however, for some people, this can be varied. A variant means that the gene has different instructions to a standard gene and as such behaves differently and can cause genetic diseases that are passed down from parent to child across generations.
Hopefully, that makes sense so far. Keep this information in the back of your mind as you read the rest of this article.
So, in order to treat genetic diseases, cell and gene therapies aim to treat, prevent and hopefully cure diseases, however, approach this problem in two different ways. Whilst this article is predominantly focused on gene therapy, I thought it would be useful to lightly define cell therapy to avoid confusion.
Broadly, cell therapy aims to treat diseases by restoring or altering certain sets of cells or by using cells to carry a therapy through the body. Cells are modified outside the body before being injected into a patient. The cells can be either a patient’s own native cells or sourced through a donor.
Gene therapy similarly aims to treat diseases by replacing, inactivating or introducing genes into cells. This can occur inside the body or outside the body.
The gene therapy value chain is quite similar to standard drug design and discovery and is as follows:
Target Identification - Similar to general drug discovery, identifying which cells to target for the highest level of efficacy is incredibly important. With gene therapy more specifically, CRISPR has been a significant breakthrough in being able to precisely cut DNA and edit parts of the genome. This has allowed for Target identification and also validation as it enables high specificity in targeting the desired gene.
Lead Optimisation and Payload Design - After the target has been identified, the next step is to figure out the best way to treat the disease by introducing genetic material into the patient's cells. This involves determining which specific gene or protein should be used. The payload can be in the form of DNA or RNA. An example of a gene therapy payload is the use of messenger RNA (mRNA) to produce a therapeutic protein in the patient's cells. mRNA is a type of RNA that carries genetic information from the DNA to the ribosome, where it is translated into a protein. In mRNA-based gene therapy, the mRNA is introduced into the patient's cells, where it is translated into the therapeutic protein. This approach has been used in the development of COVID-19 vaccines, where the mRNA payload encodes the spike protein of the virus, triggering an immune response.
Delivery Vehicle Design - Once the payload has been designed, it needs to be delivered in a stable manner and be primed for efficient integration into a genome. Most recently, Viral Vectors have emerged as the preferred delivery vehicle. Viruses are powerful delivery vehicles as they are capable of entering cells efficiently and are able to gain access to highly specific targets and hard-to-reach areas. There are currently three different vector types with adeno-associated-virus (AAV) vectors being the most commonly used and administered to patients by infusion or in vivo administration.
Translation Development - This is the preclinical development stage where GTs are tested for their safety and efficacy. In vitro and in vivo testing, toxicity and pharmacokinetic assessments are completed at this stage.
Manufacturing Development - GTs are highly personalised and require a specialised manufacturing process. Manufacturing costs for a single dose are extremely high exceeding hundreds of thousands of dollars, however, this is already down from the $1M+ range that these treatments used to cost
Clinical Trials - As with any other medical-related treatment, the final step is progressing through multiple stages of clinical trials.
As I documented in my last article about drug discovery, going through this entire process is convoluted, complex and costly. For a normal drug, it costs nearly $1B to go through this process. For GT therapies, it is multiples on that. The level of nuance required with target identification, delivery vehicles and actually manufacturing the treatment means that accessibility for these treatments is strictly reserved for the top 0.0001%.
To date, innovation in the GT space has been slower and is definitely behind that of traditional drug discovery. However, the hope is that with increasing innovations in AI and the ability to parse through vast amounts of data, this should start to accelerate exponentially over the next 5 - 10 years. CRISPR is one of the more recent breakthroughs that have improved the efficacy of treatments but also opened up new possibilities for biologists to experiment with.
Thanks for reading Superfluid! Subscribe for free to receive new posts and support my work.
Novel breakthroughs in Gene Therapy
The first attempts at gene therapy happened in the late 1980s, however, since then we’ve still been in a long initialisation phase with many years of early and failed experiments. The process is inefficient and requires lots of resources including specialised machinery in order to manufacture the therapeutics. In this section, I wanted to cover two interesting companies that are incredibly early (from a progress POV) in their journeys but are working on incredibly important parts of the value chain.
Starting off with Delivery Vehicle Design, there are numerous companies trying to improve the efficacy and manufacturability of AAV vectors. One such company is Affinia Therapeutics. Backed by Google Ventures, NEA and Atlas Venture, Affinia is fully focused on improving existing AAV-based gene therapies. They have adopted a three-prong approach:
The first part to tackle is creating capsids (the outer shell of the AAV that packages the genetic code being inserted into a patient) with improved tissue targeting, immunologic profile and manufacturability. To do this, Affinia has used ancestral sequence reconstruction (ASR) to design libraries of novel capsids. ASR is a process that involves reconstructing the evolutionary history of specific genes or proteins. In this case, Affinia reconstructs AAVs that are predicted to have been ancestors of naturally occurring capsids. This allows them to predict the protein sequences and structures of capsids that are highly effective. This process is heavily iterative and relies on large amounts of data capture and ML models.
As well as optimising the delivery of the gene, Affinia is also optimising how the body reads the genetics code for a therapeutic transgene by creating novel promoters. Their belief is that existing promoters are not optimised for certain cell types which leads to lower levels of effectiveness of the treatment. Similar to the optimisation process with capsids, Affinia has computationally designed libraries of novel promoters by using existing data on natural promoters and cell-type expression profiles. It is then an iterative process to find top performers for each distinct cell type.
Affinia ties this all together by optimising its manufacturing processes. Traditionally the manufacturing process has required a large range of equipment including bioreactors, centrifuges, chromatography columns, filtration systems, formulation tanks and sensors. It’s a capital-intensive process and the number of doses that are created in a single cycle is low (under 10). As a result, Affinia is focused on increasing the yield with each cycle which will make this a more cost-effective treatment, accessible by a larger part of the world.
We lightly touched upon the advent of CRISPR-Cas9 technology before and how that was a huge breakthrough in the field of gene therapy in 2012. Since then we’ve come further with companies such as Caribou biosciences iterating on the original CRISPR-Cas9 tech and putting their own spin on it.
For reference, the original CRISPR-Cas9 systems consist of two key molecules, 1) an enzyme called Cas9. This acts as the scissors that can cut the two strands of DNA at a specific location in the genome and 2) a piece of RNA that acts as a guide for the Cas9 enzyme to the right part of the genome. The guide RNA binds to the target and then attracts the Cas9 to cut the strands of DNA. After this, the cell either recognises that the DNA is damaged and repairs it, or scientists introduce changes to one or more genes in the genome of the cell to alter the DNA
In the case of Caribou, they discovered that there was a tendency for the RNA guide to bind with other regions of a genome that are ‘off-target’ i.e., not the original target. This means that the Cas9 enzyme would cut the wrong DNA sequence and can potentially cause harmful mutations.
In order to counteract this, Caribou created chRDNA after discovering that if they tied DNA into the backbone of an RNA guide, the bond between the guide RNA and off-target DNA would be weak, which results in a lower chance of the Cas9 enzyme cutting the wrong DNA sequence.
However, to do this practically is an arduous process. Caribou first identify specific locations within the RNA guide that need to be modified with DNA. To do this is an iterative process and can take time. There are different ways of conducting off-target detection experiments. In Caribou’s case they use an off-target detection assay (assay = lab test) called the SITE-Seq assay.
To perform the assay, they isolate genomic DNA (gDNA) from unedited cells and in a test tube place the gDNA with a Cas9 enzyme and either the pure RNA or chRDNA guide. From here, the scientists observe the differences in double-stranded-break patterns. This data is then aggregated to create a genomic map of potential off-target editing sites. The team can then use this genomic map to validate whether off-target effects actually occur at these identified sites. If so, then they need to tweak the chRDNA to avoid or neutralise those targets.
Thanks for reading Superfluid! Subscribe for free to receive new posts and support my work.
Challenges for future R&D
For both of the companies above, it’s clear to see that the processes in place to be able to test, refine and optimise the outputs that they need to produce are arduous and time-consuming.
This is the reality for basically every company that is trying to create a drug, but even more so for Cell and Gene therapy companies where the stakes are so much higher, the room for error is nil and treatments are in their infancy.
In my last article, I drew positive conclusions about AI and ML being highly appropriate for filling in gaps in key workflows for drug discovery teams, and the efficacy of conducting large-scale simulation exercises without needing to conduct physical tests. In the case of gene therapy, all of that goes out the window.
Immediately, the subject matter complexity places significant roadblocks on the pace and quality of innovation. Gene therapy is on the literally cutting edge of drug discovery. To date, all treatments have largely been once-off or in small batches. Scientists are making discoveries on a daily basis and so it is still challenging to accurately map out what solutions are meant to look like. As a result, unlike with standard drug discovery, it’s trickier to overlay AI and ML solutions as the relationships between different DNA and RNA structures and how certain combinations might behave are still being fleshed out.
Moreover, data availability is low and generating experimental data from scratch is expensive and hard given the manual processes required. As a result, it’s harder to apply AI and ML solutions immediately as training data is sparse and the risk of having false positives or compounding errors is high.
However, it must be stated that teams should have robust data collection processes as part of their operating protocols in order to harvest experimental data early. It’s clear that this will start to become a point of defensibility for companies in this space, and those with high-quality data at their fingertips will likely have a huge advantage as broad ecosystem advances are made.
Finally, from my outside perspective, gene therapy is largely reliant on wet-lab experimentation and using heavy machinery to actually produce the therapeutics. It seems like in order to actually improve the speed at which experimentation can be conducted and to initially provide a bulk of the cost savings per dose, there needs to be a significant improvement in creating special hardware or adapting existing devices to be fit for purpose in a gene therapy setting.
Without the ability to conduct experiments at a high velocity and the inability to cut down on the high-cost base, it’ll be hard for a large number of companies to even progress through clinical trials. This in turn means that it’ll be harder for the companies to harvest data, or get the feedback that is required to continually iterate and improve at their stage of the value chain, thus having a dampening effect on the speed and quality of innovation that is able to occur.
Make sure to subscribe now to not miss the next article.
Thanks for reading and see you next time!