A newer application can update a project’s copied workflow. Keep the old project as a record before you test the new release. The installed application version alone does not change a saved result.
1. Confirm the package#
Use the official release record to confirm that the intended version and platform package are available. The Windows GUI installer accepts Windows 10 version 1809 (build 17763) or later. Automatic wsl --install setup needs Windows 10 version 2004 (build 19041) or later, or Windows 11, under Microsoft’s current requirement; local analysis uses WSL2. The packaged native backend is for Linux. macOS GUI/configuration CI does not establish an installable native analysis backend. Close the application before upgrading a Windows installation.
2. Preserve the old run#
Copy the complete project to a separately named location, including config/, checks/, logs/, applicable results/ and its run summary. Record the version and route that produced each published number. Do not overwrite your only historical result or edit raw primary data to make an old run resemble a new one.
3. Reopen and validate#
On Windows, the installer updates the existing application in place without automatically running the previous uninstaller or recursively deleting its directory. Open a separate project copy with the new app, check Environment, review the selected input route and sample sheet, and run the current pre-run checks before execution. A newer version or changed bundled workflow content can refresh the project copy. Synchronization stops if the project’s workflow was edited locally rather than silently replacing that work. Review any changed status and its evidence before Start or Resume.
4. Recompute affected meta results#
Version 0.34.0 corrects pooled-effect BH adjustment: rem_padj now covers every estimable pooled test, including opposite-sign and neutral genes. Results without the corrected-method marker require recomputation before that adjusted value is interpreted. The matching-sign combined-p FDR family, meta_sig, raw pooled estimates and enrichment selection retain their definitions. Review the executed factor, numerator, denominator, study eligibility and result-call threshold in the new cross-study report; a display-only figure threshold can differ. A combined hit does not establish independent replication.
5. Compare claims, not only files#
Check whether the changed output is used by your scientific conclusion. The nine bundled biological datasets were not rerun under 0.34.0; older outputs do not validate a new run. See version notices and the provenance guide before citing a new result.