An ambitious plan to sequence genomes for 1.85 million eukaryotic species on our planet is underway. It’s a massive undertaking that will dramatically enhance our understanding of biology, and inform conservation and biodiversity restoration efforts.
AI is making this go a little bit faster, by powering tools to not only record and catalog data from DNA, but also ensure it’s accurate and free of errors across billions and billions of building blocks.
The effort I’m talking about is called the Earth BioGenome Project (EBP), a collaborative program to catalog the genomes of all of Earth’s current species over the course of a decade. It started in 2018, and is projected to cost nearly US$5 billion in total. EBP includes more than 60 global affiliated projects to record and sequence genomes, and it’s contributed data for 4,386 species thus far. That covers mammals, reptiles, fish, birds, insects, all kinds of plants, and everything in between.
This work is particularly crucial because on average, dozens of species go extinct every single day – and once they’re gone, we won’t have a chance to learn how they once lived or came to be through evolution.
Now, you’ll note that EBP has only made the tiniest of dents in its grand target of sequencing 1.85 million genomes. But bear in mind that in the past few years, AI and automation technologies have advanced rapidly, and could help accelerate efforts going forward. Google actually has an important role to play in this: its research division has developed a host of artificial intelligence tools to help speed up different parts of the sequencing process.
DeepVariant, which was released in 2018, accurately reconstructs an individual’s complete genome sequence using data generated by modern High-Throughput Sequencing (HTS) instruments from DNA. HTS machines have been around for a couple of decades are rather imperfect: they only produce billions of short DNA segments, or “reads,” rather than the full genome, so processing these reads into a single, accurate sequence is a major challenge.
DeepVariant transforms this into an image classification problem, allowing a trained deep neural network to analyze visualizations of the aligned reads and determine whether a discrepancy is a true genetic variant or merely an HTS instrument error. This technique achieves significantly greater accuracy than previous methods.
This year, the genomics team at Google Research unveiled DeepPolisher to help with genome assembly. It’s said to be capable of reducing errors in the assembly process by 50%, and prevent annotation programs – which look at raw data and identify genes and their functions – from missing disease-causing variants in genes.
With tools like these, Google Research says it can help researchers understand how they might prevent some diseases in wild species, and assist endangered species in avoiding extinction through sensitive interventions.
Decoding Life on Earth | Google and the Earth Biogenome Project
One fascinating example of this, narrated in the video above, highlights the plight of the Kākāpō. It’s a flightless parrot native to New Zealand, which can grow to be the heaviest of all parrots on Earth and live to 100 years. Sadly, only 49 individuals were accounted for in 2015.
With the genomic data of every living bird in the species, conservationists could see where their genetic diversity was greatest; that’s important for avoiding issues that arise with inbreeding, like low fertility and weak immune systems. They then translocated these birds across their habitat so they could reproduce and expand their population effectively to nearly 200 birds over the course of three generations. There are now more than 250 of them across the islands of New Zealand.
The teams involved in the EBP will have their work cut out for them in cataloguing nearly all the remaining species to meet their target by 2028. However, it should only get easier with automation and AI assistance. While sequencing the very first human genome took an entire decade and cost $3 billion up until 2003. Today, it can be wrapped up in a day at a cost of just $1,000.
Find out more about EBP on its website.
