Hybrid de novo + database search

3 methods · 2024–2026

Hybrid de novo + database search: Two engines run side by side and reconciled, a de novo sequencer and a database search, each keeping the identifications the other cannot make.

Two engines run side by side and reconciled, a de novo sequencer and a database search, each keeping the identifications the other cannot make.

The earliest of its 3 methods is Orthrus (2024); 2 more have followed.

Methods 3
Papers describing them 3
Authors 10
Active 2024-11-15 to 2026-06-29
Deep learning 2 of 3
Kinds downstream-application (2), adjacent
Acquisition DDA (3)

Methods (3)

Oldest first, by the paper that describes each one.

  • Orthrus (2024): Open-source metaproteomics pipeline combining Casanovo transformer-based de novo sequencing with Sage database search and Mokapot rescoring.
  • HDPS (2025): Heuristic two-round sequence assembly strategy combining multi-enzyme and microwave-assisted acid hydrolysis, pNovo de novo peptide sequencing, and pFind homology database search with k-mer graph assembly and majority-vote error correction; achieved 100% sequence coverage and >98% amino-acid accuracy on full-length Ricin toxin A and B chains without a reference sequence, outperforming ALPS.
  • INSearch (2026): Prototype AI-native database search that replaces the combinatorial scan with retrieval. A dual encoder projects experimental spectra and theoretical peptide sequences into one shared latent space under a contrastive objective with variance regularisation to prevent feature collapse, and an alternative alignment-and-uniformity loss; approximate nearest-neighbour search then returns the top-k candidates, cutting retrieval from the O(S × ρP) of a conventional engine to O(S(log P + K)). Because retrieval is approximate, candidates are re-ranked by InstaNovo’s decoder run in teacher-forcing mode as a scoring function, aggregating residue log-probabilities by geometric mean. On the nine-species benchmark the held-out yeast split reaches Recall@1/5/100 of 60.0/78.9/91.9%, with neural re-scoring lifting Recall@1 to 82.9%; the harder S. brodae proteome reaches only 55.5% Recall@100, pointing to a need for larger-scale training.

Applied in

Metaproteomics, Toxin identification (biodefense)

Papers describing them (3)

Authors (10)

Fuli Wang, Hao Wang, Jiale Xu, Junjie Wen, Matthew James Collins, Siyu Zhu, Tasneem Midhat Mustafa Attaallah, Yongqian Zhang, Yubo Song, Yun Chiang

Seen in the charts

Back to the full map

Back to top