Explanation: InstaNovo Performance

This document provides an overview of InstaNovo's performance on various benchmark datasets.

Performance Metrics

We evaluate the performance of InstaNovo using peptide accuracy. This metric measures the percentage of correctly predicted peptide sequences at full coverage (i.e., without any confidence filtering).

Benchmarks

We have benchmarked InstaNovo v1.1 and InstaNovo+ v1.1 against our previous models. For all results, InstaNovo decoding was performed with knapsack beam search decoding, and InstaNovo+ was used for refinement.

Nine-species dataset

This dataset contains spectra from nine different species. The models were evaluated in a zero-shot setting (i.e., without any fine-tuning on the test species).

Dataset	InstaNovo v0.1	InstaNovo+ v0.1	InstaNovo v1.1	InstaNovo+ v1.1
Bacillus	0.624	0.674	0.652	0.684
Mouse	0.466	0.490	0.524	0.542
Yeast	0.559	0.624	0.618	0.645

Biological validation datasets

We also evaluated the models on a variety of challenging biological datasets.

Dataset	InstaNovo v0.1	InstaNovo+ v0.1	InstaNovo v1.1	InstaNovo+ v1.1
HeLa degradome	0.695	0.719	0.813	0.821
HeLa single-shot	0.503	0.517	0.642	0.647
Herceptin	0.494	0.562	0.710	0.720
Immunopeptidomics	0.581	0.697	0.707	0.748
Candidatus "Scalindua brodae"	0.724	0.736	0.748	0.762
Snake venoms	0.196	0.198	0.221	0.238
Nanobodies	0.447	0.464	0.492	0.508
Wound fluids	0.225	0.229	0.354	0.364

As the results show, InstaNovo+ v1.1 consistently outperforms the previous models across all datasets.