PLMNovo

algorithm · Transformer (AR)

PLMNovo: algorithm · Transformer (AR). Casts peptide-spectrum matching as a constrained optimisation that aligns the spectrum encoder’s embeddings with those of a pre-trained protein language model, and…

Casts peptide-spectrum matching as a constrained optimisation that aligns the spectrum encoder’s embeddings with those of a pre-trained protein language model, and trains encoder and decoder with a Lagrangian primal-dual algorithm, optionally fine-tuning the language model.

Kind algorithm
Deep learning yes
Acquisition DDA
Family Transformer (AR)

Code

Live stars, open issues and last-push figures are on the Code activity chart.

Reported comparisons (2)

The comparison tables this method’s own papers print, standardised: every value on a 0-1 scale, methods down the side, the measure and then the species across. These are numbers papers report about themselves and their baselines. They are not a leaderboard, and they do not compare across tables: each was produced by a different group, on the dataset named in its corner, with each baseline either retrained, run from released weights or quoted from another paper. Where the paper says which, it follows the method’s name (hover it for the sentence); most papers do not say. Bold is the best value in a column and underline the runner-up, our ranking rather than the paper’s own marks.

Table 1

Protein Language Model-Aligned Spectra Embeddings for De Novo Peptide Sequencing, page 6: Amino acid-level and peptide-level performance comparison on the MSKB and multi-enzyme test sets [32]. Numbers in bold represent the best result under each (metric, dataset) combination.

a dataset per row group Amino acid precision Amino acid recall Peptide precision
MSKB (tryptic) Multi-enzyme (non-tryptic) MSKB (tryptic) Multi-enzyme (non-tryptic) MSKB (tryptic) Multi-enzyme (non-tryptic)
Casanovo v4.2 · retrained 0.8919 0.5200 0.8884 0.5211 0.7381 0.2467
PLMNovo ESM-2 8M 0.8973 0.5249 0.8921 0.5283 0.7411 0.2410
PLMNovo ESM-2 650M 0.8976 0.5356 0.8915 0.5323 0.7400 0.2483

MSKB is MassIVE-KB, under the train/test split (200,000 tryptic test spectra) published with Melendez et al.'s enzyme-bias data on Zenodo 12587317, the paper's reference [32]; that split mixes MassIVE-KB v1 and v2.0.15, so no version is recorded. The multi-enzyme rows are the held-out non-tryptic set from the same deposit.

Table 2

Protein Language Model-Aligned Spectra Embeddings for De Novo Peptide Sequencing, page 7: Performance comparison results in terms of amino acid precision and peptide recall on the ninespecies dataset [55]. The performance of the baseline methods, except for Casanovo v4.2, is reported from [65]. Bolded and underlined results represent the first and second best performance per species (or averaged across species in the last column), respectively.

Nine-species benchmark Amino acid precision
Bacillus subtilis Candidatus Thiodiazotropha endoloripes Apis mellifera Homo sapiens Methanosarcina mazei Mus musculus Vigna mungo Solanum lycopersicum Saccharomyces cerevisiae Average
DeepNovo 0.742 0.602 0.630 0.610 0.694 0.623 0.679 0.731 0.750 0.673
PointNovo 0.768 0.589 0.644 0.606 0.712 0.626 0.730 0.733 0.779 0.687
Casanovo · retrained 0.749 0.603 0.629 0.586 0.679 0.689 0.668 0.721 0.684 0.667
AdaNovo 0.739 0.642 0.650 0.618 0.728 0.646 0.719 0.740 0.793 0.697
Casanovo V2 · retrained 0.790 0.681 0.706 0.676 0.755 0.760 0.748 0.785 0.752 0.739
Casanovo v4.2 · retrained 0.793 0.678 0.705 0.668 0.687 0.756 0.753 0.791 0.768 0.733
PLMNovo ESM-2 8M 0.794 0.684 0.707 0.670 0.755 0.759 0.764 0.790 0.755 0.742
PLMNovo ESM-2 650M 0.797 0.688 0.709 0.676 0.755 0.760 0.766 0.795 0.771 0.746
Nine-species benchmark Peptide recall
Bacillus subtilis Candidatus Thiodiazotropha endoloripes Apis mellifera Homo sapiens Methanosarcina mazei Mus musculus Vigna mungo Solanum lycopersicum Saccharomyces cerevisiae Average
DeepNovo 0.449 0.253 0.330 0.293 0.422 0.286 0.436 0.454 0.462 0.376
PointNovo 0.518 0.298 0.396 0.351 0.478 0.355 0.511 0.513 0.534 0.439
Casanovo · retrained 0.537 0.330 0.406 0.341 0.478 0.426 0.506 0.521 0.490 0.448
AdaNovo 0.528 0.372 0.431 0.373 0.496 0.467 0.546 0.530 0.593 0.481
Casanovo V2 · retrained 0.622 0.446 0.493 0.446 0.557 0.483 0.589 0.618 0.599 0.539
Casanovo v4.2 · retrained 0.603 0.421 0.478 0.437 0.498 0.468 0.558 0.608 0.584 0.517
PLMNovo ESM-2 8M 0.602 0.423 0.484 0.434 0.544 0.470 0.565 0.606 0.571 0.522
PLMNovo ESM-2 650M 0.601 0.425 0.483 0.432 0.544 0.464 0.576 0.606 0.579 0.523

Paper describing it

Authors (4)

Navid NaderiAlizadeh, Christian Dallago, Erik J. Soderblom, Scott H. Soderling

Seen in the charts

Back to the full map

Back to top