#307: Prime Editing Could Accelerate The Adoption Of Clinical Genetic Testing, & More

While the DNA of all human beings is 99.9 identical, the remaining 0.1% includes genetic variants that determine our unique traits as well as our predispositions to disease. Genetic variants ranging from small misspellings to large chunks of missing DNA code are scattered throughout our genomes. Hereditary genetic testing aims to measure those variants and determine whether or not they are safe (benign) or dangerous (pathogenic).

Because human genomes are large and complex, clinical labs often surface mutations known as variants of uncertain significance (VUSs) that leave physicians and patients with many unanswered questions. As a result, clinical labs are focused on minimizing VUS.

To classify genetic variants, labs ask a series of questions. Is the patient’s genetic variant in internal or external databases? What does scientific literature suggest about the variant?

In recent years, some labs have used biomolecular computer models to simulate the downstream effects of genetic variants, the most advanced using in vitro gene-editing techniques like CRISPR/Cas9 that recreate a patient’s genetic variant(s). This functional testing can provide significant answers but involves risks like off-target effects and lower editing efficiency. Functional testing is a key area of research in large genetic testing labs.

Recently, a Canadian research group applied one of the newest forms of gene editing, “prime editing,” to improve functional testing for genetic variant interpretation. In that study, prime editing improved editing efficiency, edited the genome in places CRISPR/Cas9 could not, and enabled cheaper, faster, and more scalable functional testing.

New genome editing tools like prime editing are likely to enable better functional tests that help labs contend with VUS, accelerating the adoption of clinical genetic testing.

Large language models are outperforming their smaller predecessors across a wide range of tasks, demonstrating that size matters, particularly in vision models. Last year, Facebook AI (now Meta AI) unveiled SEER, a 1 billlion-parameter computer vision model that learns from random, unlabeled, and noncurated sets of images. SEER was an exciting advancement because “self-learning” is more scalable than the prior computer vision models trained on human-labeled datasets. Recently, Meta AI scaled the SEER model to 10 billion parameters, increasing its performance dramatically.

SEER’s 10B-parameter model delivered fewer biases than other models trained on highly curated datasets, suggesting the intriguing possibility that larger models will reduce bias, a conclusion contrary to the conventional wisdom that human labeling and curation are the only––or best––techniques for decreasing bias.

The global upheaval caused by Russia’s invasion of Ukraine is demonstrating the power of cryptoassets as a source of economic independence. A week after its plea for crypto donations in a tweet, for example, the Ukrainian government had raised over $54 million in bitcoin, ether, and other cryptoassets. Simultaneously, Russian citizens turned to bitcoin as a hedge against the rapid devaluation of the ruble, pushing BTC-RUB trading volume up by 243% in the last week of February. With the rapid devaluation of the ruble, bitcoin’s market capitalization is more valuable today than the Russian money supply.

While some critics have suggested that Russia is using cryptocurrency to circumvent unprecedented sanctions, supporters disagree, claiming that “liquidity in crypto is still a fraction of the global currency market.” Average income earners in Russia might be trying to preserve their purchasing power and asset base, but oligarchs would need much more that the roughly $1 trillion crypto market capitalization to do the same. In addition, Bitcoin blockchain’s transparency probably would expose any “whale” activity.

Meanwhile, controversy at blockchain infrastructure provider Infura and NFT marketplace OpenSea during the last few weeks have highlighted that dependence on any central service provider could undermine the benefits of crypto networks. To comply with sanctions, Infura and OpenSea had to ban IP addresses originating from Russian controlled regions of Ukraine, including Donetsk and Crimea, and delist individual accounts. Users who self-custody private keys through digital wallets like MetaMask also suffered because they rely on Infura for connections to the Ethereum network. While some tech-savvy users circumvented regulatory measures with VPNs and blockchain nodes, the average user could not make any moves in crypto.

In our view, crypto networks that rely on central service providers have opted for ease of use and a single point of failure over decentralization. While they can be convenient in ideal, everyday scenarios, centralized service providers often cannot cope with extraordinary disruptions which, unfortunately, seem to be more ordinary these days. As a result, companies like Block that are focused on developing privacy and decentralization for the custody and exchange of bitcoin are likely to gain share and thrive.