Hi all, Following up from Keiran, we had a meeting on our side and came up with the following priorities for a release roadmap: V2.1.0 (Theme: QoL changes): • Cherry pick everything except the partially complete extract metrics refactor from hackathon branch and merge to dev: ◦ Auto monomer/multimer mode detection (PR #527) ◦ meta.yaml additions (PRS #537 and #560) • Review PR #556 – real test (need to see how this can be managed with runners). • Issue #586 – error with local msa using non-protein entities. • UNSW ackowledgement (PR #573) V3.0.0 (Theme: Update metrics handling): • Deprecate HelixFold3, RF-AA, AlphaFold2 (standard – ie not split), AlphaFold2 (monomer preset – the idea being to only support ptm for monomers), RF2NA. This might be controversial so happy to discuss this in a development meeting. My opinion is that these modes do not represent current best practice and add disproportionate maintenance burden for ongoing development. All downstream work will be simpler if we can deprecate these tools. • Refactor plotting code (PR #574) • Refactor extract metrics. We have discussed this extensively and have come up with the following solution: ◦ Use a single EXTRACT_METRICS module which handles all of the modes within 1 python script. ◦ Provide a blob of data to this module (essentially our current raw directory) and let the python script check for the necessary files. ◦ We can alias each mode to provide the container used to generate the data (ie some modes need jax/torch to parse raw output data). • Calculate ipSAE scores for multimer predictions (Issue #517). • Combine output metrics in a modelcif file (Issue #578). This will let it natively be parsed by tools like ChimeraX to visualize structure and PAE together. • Add support for multiqc to better capture sample metrics in a bulk run (PR #450) • Review any missing version tracking in versions.yaml (database versions etc). V3.1.0 (Theme: enhanced features): • Batch ColabFold Search to allow much more efficient bulk runs (PR #551) • ColabFold Search GPU to accelerate large scale runs (PR #497) • Add protenix v1 mode (PR #559) Miscellaneous features (would be good to get in - some are blocked by various issues): • Update HTML report to use PAE-viewer to allow interactivity between structure and PAE (Issue #341) • AF3 split mode (Issue #577). • Better support for multiple replicates/sampling (Issue #478). • Fine-grained saving of intermediate results (discussed in PR #453). • Single canonical output across methods (Issue #339).