Today we announced our flagship project- a $60k challenge to predict the designer of genetically engineered sequences.

Biologists, statisticians, deep learning folk- we need your skillz!!

So what the heck is attribution, and why do I think you should care? (a thread) 1/n https://twitter.com/altLabs_tech/status/1295733200187260930

First, some background. Bioengineering is growing crazy fast. If you haven't seen this, that blue curve is moore's law compared to DNA reading and writing productivity. DNA sequencing is growing faster than the most famous law of speedy technology (2/n) ( https://synbiobeta.com/time-new-dna-synthesis-sequencing-cost-curves-rob-carlson/)

As tech develops, so does its capacity to have large consequences on the world. Biotechnology in particular touches the most critical aspects of our lives- our food, ecosystem, and bodies. Now, as we all wait for a bit of biotech we call a vaccine, this is especially clear 3/n

I am pumped about the potential for biotech to improve our lives and contribute to addressing global challenges. IF we take on the charge of doing so responsibly. That includes: 4/n

Do choices in tech development exacerbate or reduce inequality? Are their unintended side effects which might pose a risk to human and environmental health? Are developers of technology accountable to the communities who will be affected by their work? (5/n)

As a researcher, I often find it frustrating to think about how my projects will affect the world because of how distant it feels from the day-to-day work.

This is different. We think we have found a problem that, if we solve it, can directly enable responsible biotech.6/n

It's called genetic engineering attribution (mouthful ikr). This is what it boils down to: if I give you an organism and tell you its genome has been tinkered with in a lab, can you look at the DNA sequence and a few experimental measurements and guess which lab designed it? 7/n

For biologists: genetic engineering leaves "clues" as to the designer. I have a favorite cloning method, favorite restriction sites, go-to primers, that-one-linker-which-works, and I typically work on protein engineering projects.

Together these form a genetic "signature". 8/n

For CS folk: we have a well-defined supervised classification problem (50k+ examples) from (DNA sequence, experimental metadata) -> Laboratory ID. This is a playground for biological-sequence language models, multimodal learning, and challenges for calibration/ interpretation 9/n

But why does this matter? As things currently stand, we have no systematic way of establishing a link between the "code" an engineer "wrote" and the designer themselves. This makes it hard to assign due credit to the hard work that went into a genetic system. 10/n

More importantly, it means that genetic engineering *lacks transparency*. Not for lack of effort towards open science, but because you can't just drop a "Made by me" in the DNA like you could in the README of your github project. 11/n

And so, if there comes a time when we need to be better- a failure of biosafety protocol, say, or an unintended consequence affecting a community-we *won't know where to start*. We can't hold ourselves and others accountable if the results of biotech are anonymous by default 12/n

I don't want to neglect the large body of work in a field called microbial forensics that proceeds this- so far, despite its cost, that has done the trick. But remember those exponential curves! Part of being proactive is building better tools *before* we need them. 13/n

We need your help to improve attribution tools. We're offering $60k in prizes to teams who can improve attribution accuracy OR make innovative contributions on the difficult-to-quantify real-world attribution challenges. 14/n

We have an incredible panel of experts across a huge span of disciplines including @kesvelt @geochurch @T_Inglesby @ggronvall @DrewEndy @DavidRelman @RosieCampbell @jdiggans and more. 15/n

Finally, an enormous thanks to the team @altLabs_tech and our partners at John's Hopkins Center for Health Security, the Applied Physics Laboratory, and iGEM Safety and Security Committee.

Now, go win! https://altlabs.tech/geac/ 

(I know he keeps a low twitter profile but @willbradbio is amazing and deserves the credit for making this happen!)