Although the bug is still valid and it is impossible to compile cerebro using nodejs v9, I decided to add nodejs v8 to makedepends, so that those of you who build AUR packages in chroot (e.g. via aurutils) will be able to upgrade.
This is certainly possible, but I'd suggest creating a new AUR package cerebro-bin. The convention is to name AUR packages cerebro if we build the app from sources, and cerebro-bin if we use pre-compiled binaries (such as deb package).
However, talking specifically about cerebro, I must say I've seen more than one case when pre-compiled packages were not based on the release tarball. I'm not exactly sure how that happened, I believe this was a honest mistake, but again - it happened more than once.
We use a server with RHEL7 to run CEREBRO and this is what we have done to try to fix all these administrative shortcomings without using so much time. Maybe you will have to adjust some of the steps if you are using another linux distribution:
Schematic workflow of Cerebro. In the first step, the raw data are processed and analyzed (barcode extraction, alignment, etc.) using existing tools such as Cell Ranger, stored in a Seurat object, and exported to a .crb file using functions of the cerebroApp package. Subsequently, the .crb file can then be loaded into Cerebro for visualization. Currently, Cerebro can be launched as a standalone application, from the cerebroApp R package, or from the dedicated Docker container