A fully automated benchmarking suite to compare macromolecular complexes - Nature Methods This study advances methods for benchmarking macromolecular complex predictions by introducing a scalable open-source framework used in recent community assessments to compare structures, interfaces a...

And the full DOI link is here: doi.org/10.1038/s415...

🚀 New paper in @natmethods.nature.com!
We present OpenStructure's powerful scoring capabilities, used to assess predictionsin CAMEO and CASP.
Read the full study here:
🔗 doi.org/10.1038/s415...
#StructuralBiology #Bioinformatics #OpenStructure #CASP #CAMEO #ProteinStructure

The future of structure prediction benchmarking: measuring progress and breakthroughs · Luma Benchmarking has been a key driver of progress in protein structure prediction methods. As the field continues to evolve, several key questions prevail: How…

🔬 Workshop: Future of Structure Prediction Benchmarking
📅 Sept 8, 2025 | Basel
💡 Talks + breakout sessions on #CASP #CAPRI #CAMEO & benchmarking for drug discovery
🎟️ Free registration (limited spots): lu.ma/ws9nu1xf
Join us to explore how benchmarking can drive breakthroughs in structure prediction.

Figure 4 in the Runs N' Poses paper shows how one can combine sequence identity with ligand similarity metrics to assess the structural novelty of protein-ligand complexes.

Runs N' Poses is a great example of how to rigorously filter protein–ligand complexes for structural similarity: www.biorxiv.org/content/10.1....

💡Note: these are sequences where the only template available is < 40% sequence identity, making it already a stricter scenario than usual splits with 2021 training cutoffs.

Even at 40% sequence identity, SWISS-MODEL models have an LDDT greater than 0.7 on average. If you want to build a meaningful validation set without leakage, you need to adopt a significantly more stringent criterion.

We looked at 3891 past CAMEO targets (January 2020 to April 2024). We plotted the LDDT of homology-based SWISS-MODEL prediction against the sequence identity of the top BLAST template, filtering for hits with at least 80% coverage of the target sequence.

Scatter plot of LDDT as a function of sequence identity for high coverage homology models. The horizontal red line at 40% sequence identity highlights the presence high quality models in the low sequence identity region.

Filtering out homologous structures from the PDB at 40% sequence identity is not enough to create a robust test set. Significant leakage persists at this level, and comparative modeling can still produce high quality models.

Create your account and register your server now!
cameo3d.org

The target page with server results and superposed 3D structures.

Detailed per-target scores and interactive 3D visualization.

The CAMEO results page with custom timeframes, common subset comparison, and boxplots

Custom timeframes, automatic server comparisons on shared targets, multi-metric evaluations, and interactive boxplots for selected servers.

Introducing CAMEO Structures & Complexes - automated weekly blind benchmarking of structure prediction servers. Now with heteromeric and protein-ligand complexes.
Join us and register your server now!
cameo3d.org