Work on it
Start with one small thing → — four contributions written as numbered steps, each finishable today. Everything here is open, and most of it is unfinished. This is not a project looking for contributors so that it can say it has contributors — these are jobs that are actually blocking things, written honestly, including how big each one is and what it needs from you.
The shape of the work is unusual and worth understanding before you pick something. Almost nothing here required new laboratory data. The genome was published in 2017, the developmental transcriptome in 2016, and the results we are proudest of came from putting them next to each other and asking a question that neither paper asked.
So the most valuable thing you can bring is not equipment. It is a field you know that we do not. The best result in this project so far came from computational geometry, a hundred miles from biology, and it took someone noticing that two problems had the same shape.
If you write code
Take a gene family and place an unknown gene in it
We have a tool that does this — domain search, alignment, maximum likelihood, and a report of which reference clade a query falls into with bootstrap support. It currently knows one family. There are several dozen unnamed proteins waiting, each of which becomes a named gene in the graph if it works.
Step-by-step: name one protein →
needs: Python, patience with bioinformatics file formats. No biology required — the tool explains what it is doing.
The graph has relations it cannot express
Sixty-seven biological processes have 907 claims pointing at them and almost no relations leaving them. The knowledge is sitting in claim prose — "moulting requires ecdysone", "gastrulation precedes segmentation" — and the schema has no edge for it. There is a tool that finds these candidate relations by pattern; turning its output into reviewed edges is open-ended, high-value, and does not need a biologist to start.
Step-by-step: check one claim →
needs: Python, or just careful reading and a text editor.
Make the simulation wrong in an interesting way
There is a cell-level embryo simulation that runs in a browser from a seed. Its rules are nodes in the graph pointing at functions in the code, and every comparison with real data is a test node that records pass, partial or fail in words. Several fail. Finding a new way to break it against published observations is a contribution, and a failing test is a result we keep.
Step-by-step: tell us something is wrong →
needs: JavaScript, and a willingness to read a staging table.
If you have institutional access
Roughly a hundred papers we cannot legally reach
We do not get around paywalls, logins or bot checks. So there is a list of papers and supplementary files — several of which are the single thing blocking a preprint — that sit there until someone with a university login can open them. One of them is a 1972 German ultrastructural study that every modern paper argues with second-hand.
The fetch list is public — take any one →
needs: a library card, essentially.
If you have a microscope, or a garden
Go and find them
Tardigrades live in moss and lichen almost everywhere, including on the roof of wherever you are reading this. Soaking a patch of moss in water for a few hours and looking at the squeezings under a cheap microscope works. There is a page on finding and keeping them. Records with a location, a date and a photograph are genuinely useful: the distribution data for most species is thin, and several of the papers we have just filed are exactly this kind of survey.
Step-by-step: find your first tardigrade →
needs: moss, a dish, and a microscope that costs less than a phone.
Imaging we cannot do
Confocal stacks, SEM, and anything with a fluorescent antibody against a neural marker. We have built the analysis side and have far more analysis than data. If you have access to an instrument and an hour of idle time on it, tell us and we will tell you exactly what would be most useful and why.
Not sure it fits? Just write and ask →
needs: an instrument, and someone who knows how to drive it.
If you know a field we do not
Tell us our problem is your problem
This is the one we most want and can least specify. If something here looks like a thing you already know how to solve — a packing problem, a signal processing question, an inference method, a rendering trick, a legal or archival question — say so. The transfer of a method from one field to another is not a lesser contribution than inventing it. It is usually the harder half, because it requires someone who can see both sides, and those people are rare.
No steps for this one — just write →
needs: your actual expertise, and ten minutes.
How to start
The repository holds everything: the graph, the tools, the sources manifest, the simulation and this website. Read how it is built first, and the engineering page if you are technical. The open questions are real and most of them are not ours — they are questions the papers themselves posed and nobody went back to.
Then get in touch at meow@meow-meow.io. Say what you would like to work on, or ask what would help most. There is no onboarding process and no contributor agreement; there is a person who will write back.
The project is run by Meow-Ludo Meow-Meow in Canberra, Australia. The graph and the code are CC BY 4.0 and MIT respectively; the facts belong to the sources cited on every claim.