If you record the same concert with two different microphones, the hiss in the recording will differ. But if the same violinist plays a note a little louder or a little softer each time, both microphones will pick that up. Cells in one person, all carrying the same genes, switch those genes on with varying strength in much the same way. This is called transcriptional noise. Scientists wanted to know: is this noise a property of the genes themselves, or an artifact of the instrument used to measure gene activity?
To find out, they took human cells (the GM12878 line) and two popular instruments for reading RNA: 10x Chromium and Parse Evercode WT. Each instrument was used four times, as if the same concert had been recorded four times. It turned out that the instruments capture RNA molecules differently: one is better at seeing RNA with a tail of adenines, the other at seeing RNA without one. But when the noise itself was measured, that is, how far a gene's activity deviates from the average, the results matched. Genes that were noisy in the data from the first instrument were noisy in the data from the second as well. What's more, the noise turned out to be stronger for long genes and weaker for genes with a CpG island, a stretch of DNA where the letter pairs C and G occur frequently.
Now it's clear that transcriptional noise is not a defect of a particular instrument but a feature of the gene itself. That means data from different labs can be pooled and noise studied together. Admittedly, why long genes are noisier and CpG islands dampen noise is still unknown. But that's a question for next time. The main thing is that we've learned to tell the cell's music from the microphone's hiss.
The Parse Evercode WT platform uses combinatorial barcoding: cells are sequentially mixed and split, receiving unique combinations of barcodes, as if each cell gets its own postal code.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.