Computational biology · Industrial biomanufacturing

20n

A Y Combinator biotech whose software invented a fermentation route to the active ingredient in Tylenol, then published its whole stack as open source in March 2017 and went quiet. The code is still on GitHub. The company is gone.

Four vials on a bench

The photograph on the company blog shows four vials: sugar, broth from an engineered yeast, clarified supernatant, and a heap of white powder. The powder is acetaminophen, the active ingredient in Tylenol.

Until then, every gram of acetaminophen came from petroleum. The molecule entered medicine in 1955 and had never been made by a cell.

What makes the vials interesting is how the route was found. Nobody at 20n set out to make acetaminophen. The company had written software that enumerated every molecule reachable by inserting genes into a cell, and acetaminophen fell out of the list. As the blog put it: “The pathway came entirely through the software predictions, with no human input.”

The recipe was a mushroom gene, 4ABH, plus over-expression of two genes the cell already had. “Our software predicted that adding only a single gene, from the common grocery store mushroom, would enable E. coli to produce this common headache medicine,” co-founder Saurabh Srivastava told Motherboard in 2015. Then: “When we tried it, it worked.”

The problem: biology had no compiler

Specialty chemicals are a very large business made by an ugly process. Petrochemical synthesis is energy hungry and, for some molecules, impossible. Fermentation is the obvious alternative, and the field had spent a decade proving that a microbe could be rewired to make a target compound.

The bottleneck was not the wet lab. It was knowing which genes to put in. Finding a route to one molecule was a multi-year project done by hand. 20n argued that this was a search problem, and that search problems belong to computers. Its pitch: data mining could reach roughly 100 times more chemicals than were previously thought possible.

The company put that ambition in writing: “In an age where software startups can get to worldwide scale with a few engineers, we ask if biotech can do the same?” It owned no lab. Predictions went out to contract research organisations and to automated cloud labs such as Transcriptic and Emerald Cloud Lab, and the data came back.

A professor on leave and a program synthesis postdoc

J. Christopher Anderson was a UC Berkeley bioengineering professor who had worked in synthetic biology for seventeen years. Saurabh Srivastava had a computer science PhD from Maryland on satisfiability-based program synthesis, the art of getting a machine to write code from a specification, and was a postdoc at Berkeley.

They met at Berkeley and started building the software in 2011, inside a DARPA-backed lab. The culture gap was real. “We had to figure out a language to work in that both a biologist and computer scientist could understand,” Anderson told TechCrunch. Srivastava put a number on it: “With such different backgrounds, we spent an entire year simply developing a common vocabulary to be able to discuss this challenging interface.”

UC Berkeley records the company as founded in 2013 and lists ten inventors on the underlying university work, including Anderson, Srivastava, Rastislav Bodik and Sanjit Seshia. By the time 20n was hiring, the split was formal: Srivastava was CEO, and, per the jobs page, “Chris (CSO) was a Bioengineering professor at UC Berkeley before going on leave to work on 20n full time.”

The name is a joke for chemists. Twenty is the number of standard amino acids, n is the length of a protein, and 20 to the power n is the space the company said it got to play with.

DARPA money, then the code goes public

The federal record is the clearest part of the story. On May 7, 2014, DARPA signed contract HR001114C0076 with 20N LABS, INC. of San Francisco: $699,453, firm fixed price, for Living Foundries: 1000 Molecules. A second, HR001115C0152, followed on September 9, 2015 for $612,026.

That put a twelve-person startup on a contractor list with Zymergen, Amyris, Duke and two large state universities. 20n went through Y Combinator in Winter 2015 and made TechCrunch’s honourable mentions for the batch.

In January 2016 it filed three provisional patent applications, converted a year later into US applications on acetaminophen, on carnosine and beta-alanine, and on choline. In October 2016 it published a long technical post on using deep learning to read untargeted mass spectrometry data, which reached the front page of Hacker News.

Then, on March 23, 2017, the entire stack went to GitHub under GPL v3. “Over the last 4 years we have developed a better way to bioengineer organisms. We are now open sourcing our entire software stack, 20n/act.” The last commit landed four days later. Nothing has been pushed since.

What is proven, and what is still claimed

Read plainly: the engineering happened and the government paid for much of it. What never arrived was the second act. There is no publication, no named commercial contract, no product still sold, no equity filing at the SEC. The strongest surviving artefact is the source code. The strongest signal about the ending is a patent assignment: in October 2017 the company handed its three inventions back to its chief executive as a private individual.

What to watch

• Whether anyone revives 20n/act. It has sat unchanged since March 27, 2017, but it is GPL v3, and the retro-biosynthesis field has grown around the problem it attacked.
• The abandoned applications. All three published in July 2017, so the acetaminophen route is permanent prior art available to everyone.
• Whether the acetaminophen result is reproduced in the literature. Nine years on, no peer-reviewed paper carries the company’s name.
• Where the people went. Srivastava founded a second YC company, led code research at Essential AI, and now works on Nemotron at NVIDIA. Anderson teaches synthetic biology and biodesign automation at Berkeley again.

In their words

Related companies

• Ansa Biotechnologies: Also in Industrial Bio. An enzymatic DNA synthesis company in Emeryville that sells the genes other vendors decline
• Synvivia: Also in Industrial Bio. A UC Berkeley spinout that built molecular on and off switches so engineered microbes die unless they are fed a specific chemical
• Kern Systems: Also in Industrial Bio. A Boston company spun out of George Church’s Harvard lab to write digital data into DNA with enzymes instead of chemistry
• Lygos: Also in Industrial Bio. A Berkeley Lab spinout that taught yeast to make malonic acid without cyanide, then became a specialty chemicals supplier whose polymers end up in shampoo, diapers and

Sources

1. Public recordFederal contract HR001114C0076, DARPA to 20N LABS, INC., $699,453USAspending.gov · signed May 7, 2014
2. Public recordFederal contract HR001115C0152, DARPA to 20N LABS, INC., $612,026USAspending.gov · signed Sep 9, 2015
3. Public recordUS patent applications 2017/0211103, 2017/0211104 and 2017/0211105USPTO · published Jul 27, 2017
4. Public recordRepository 20n/act, metadata, contributors and READMEGitHub · created Mar 23, 2017, checked Sep 23, 2026
5. Public record20n Labs, Inc. technology listing with inventors and founding yearUC Berkeley IPIRA
6. Independent20n, A YC Synthetic Biology Startup, Uses Software To Engineer Microbes For Chemical-MakingTechCrunch, Kim-Mai Cutler · Feb 26, 2015
7. IndependentSoftware Is Designing Useful Microbes That Don’t Exist in NatureMotherboard, VICE, Kari Paul · Mar 26, 2015
8. IndependentDARPA works with six organizations on biological materialsMilitary & Aerospace Electronics · Oct 1, 2015
9. Independent20n Labs company profileCB Insights · accessed Sep 23, 2026
10. Company20n blog: open sourcing 20n/act, the acetaminophen result, deep learning for LCMS20n.com, Internet Archive · 2016 to 2018
11. Company20n open positions page, with founder roles and cloud lab vendors20n.com, Internet Archive · Jun 15, 2015
12. FounderSaurabh Srivastava, personal site, timeline and patent listsaurabh-srivastava.github.io · accessed Sep 23, 2026