How it works

Cancer genomics datasets are large, and the calculations over them are repetitive. CureCrunch divides that work into small units that an ordinary phone or laptop finishes in seconds.

1. Splitting the data

We start from open-access studies of The Cancer Genome Atlas (TCGA), which describe tumors from many cancer types. For each study and each kind of analysis, the data is cut into work units. A unit holds up to 100 tumor samples, or a block of genes across those samples, so that one unit is small enough to download quickly and compute in a few seconds.

Every unit has a fingerprint (a SHA-256 checksum). Each result is submitted with the fingerprint of the unit it was computed from, and the server rejects a result whose fingerprint does not match.

2. Computing on your device

Your device asks the CureCrunch server for a unit, downloads it, and runs the analysis locally. The analysis is a small fixed program, not code that changes from unit to unit. We are moving every CureCrunch app to a single sandboxed WebAssembly program, so a Mac, an iPhone, an Android phone and a Windows PC all run byte-for-byte the same calculation.

Only a short summary leaves your device: for example, the number of samples analyzed and the genes that stood out. The unit's data is not sent back.

3. Checking on a second device

The same unit goes to a second device. The server compares the two summaries field by field. Numbers may differ by at most 5%, lists such as gene names must match, and at least 90% of the fields must agree. If they do, the result is marked verified. If they do not, a third device computes the unit and the majority decides. A unit that still has no majority is held for review and is not counted.

This is how CureCrunch limits the effect of a faulty device, an out-of-date app or a bug on a single machine.

4. Combining the results

Verified results are pooled into summaries for each study. For example, the share of tumors in a cancer type with a mutation in a given gene, or the median mutational burden. We are writing the pooling steps up as public methods, so that statisticians and cancer researchers can review them.

What volunteers receive

Each completed unit earns points on your device's profile. We use points to recognize contributions, and they have no cash value.

Want to take part? Volunteers join in small groups so we can watch each stage closely. Join the waitlist and we will email you when a place opens.