Home · Benchmark Report

MolDyn Benchmark Report

Public distribution · v1 · 2026-04 · Protease-focused virtual screening

Executive Summary

MolDyn v1.0 is a proprietary Wave-Resonance virtual screening platform. This report summarises external validation on the community-standard DUD-E benchmark (Mysinger et al., J. Med. Chem. 2012) and reports library-scale speed measurements.

Headline results (DUD-E): AUROC 0.782 on HIV-1 protease (EF@1% = 12.66) and AUROC 0.744 on thrombin (EF@1% = 6.63) — both inside the published band of industry-standard docking methods.
Kinase envelope: on the three DUD-E kinase targets (AKT1, VEGFR2, EGFR) AUROC is 0.42–0.53 — a known v1.0 envelope limitation, addressed in v1.1 (ETA 2–4 weeks).
Speed: 16.77 ms per query on a 2.6M-molecule library; query latency is independent of library size once the one-time library build is complete.
Validated antiviral case: EF@1% = 50 on SARS-CoV-2 Mpro (PDB 6LU7) — single-target validated case study.

MolDyn v1.0 is recommended for aspartic, serine, cysteine and viral protease targets. Kinase programs should wait for v1.1. GPCRs and nuclear receptors are not yet characterised.

Methodology

Input: SMILES strings. Output: ranked similarity scores.

Dataset: DUD-E (Database of Useful Decoys: Enhanced), Mysinger et al. 2012 — the community-standard benchmark for ligand-based virtual screening. For each target, actives are mixed with a property-matched decoy set approximately 50× larger. Methods must distinguish actives from topologically distinct but physically similar decoys.

For each target, a small set of known actives is used to form a query; the query molecules are held out from the library during scoring. The library is ranked by similarity score and evaluated against ground-truth active / decoy labels. Metrics reported: AUROC (area under ROC curve), EF@x% (enrichment factor in top x%), and a one-sided rank-test p-value.

Architectural disclosure. The internal architecture of the screening engine is proprietary and not disclosed in this report.

Results — DUD-E 5-target panel

Target	Class	AUROC	EF@1%	EF@5%	p-value
HIV-1 PR	protease (aspartic)	0.782	12.66	7.65	7.5 × 10⁻⁸⁰
Thrombin	protease (serine)	0.744	6.63	4.52	3.9 × 10⁻⁶⁹
AKT1	kinase (v1.0 limitation)	0.527	0.70	0.70	5.6 × 10⁻²
VEGFR2	kinase (v1.0 limitation)	0.494	0.50	0.70	6.5 × 10⁻¹
EGFR	kinase (v1.0 limitation)	0.422	0.00	0.63	1.000

Protease targets (top two rows) demonstrate strong discrimination: p-values far below any reasonable significance threshold, AUROC inside the published band of enterprise docking methods. Kinase targets (bottom three rows) fall outside the v1.0 envelope and are documented here for transparency; kinase coverage is being addressed in the v1.1 update.

Speed Benchmark

Metric	Value
Core search time	16.77 ms on a 2.6M-molecule library (pure GPU waves × query)
End-to-end HTTP latency	~280 ms per call (adds SMILES parse + fingerprint + FFT + audit log)
Throughput	O(1) — query time independent of library size (after the one-time library build)
Library build	One-time pass over the library; queries served at the above latency thereafter
Library scale validated	up to 2.6M molecules

Comparison with Industry Baselines

Method	HIV-1 PR AUROC	Speed
MolDyn v1.0	0.782	16.77 ms on 2.6M molecules
Glide SP	~0.75	hours per 1K molecules
AutoDock Vina	~0.65	days per 1K molecules

MolDyn HIV-1 PR AUROC is within the published range of enterprise docking (Glide SP band 0.70–0.82) while delivering library-scale screening several orders of magnitude faster. Industry AUROC ranges are taken from the standard literature for DUD-E targets.

Known Limitations

MolDyn v1.0 is optimised for protease targets. Direct DUD-E validation on HIV-1 protease (aspartic) and thrombin (serine); a single-target validated case on SARS-CoV-2 Mpro (PDB 6LU7). The deployment envelope extends to aspartic, serine, cysteine and viral proteases in general (HIV-1 PR, HCV NS3, SARS-CoV-2 Mpro, dengue NS2B-NS3, thrombin, factor Xa, plasmepsin, ClpP, cathepsins, renin).

Kinase targets (AKT1, VEGFR2, EGFR in the DUD-E panel) fall outside the v1.0 envelope. Kinase support is in development and will be released in v1.1, with re-run DUD-E kinase benchmarks using identical methodology. ETA: 2–4 weeks.

GPCRs and nuclear receptors are not yet characterised on DUD-E and are outside the current scope.

Scope of Use

Suitable for v1.0	Not recommended until v1.1
Antiviral protease programs — HIV-1 PR, HCV NS3, SARS-CoV-2 Mpro, dengue NS2B-NS3	Kinase programs — AKT, VEGFR, EGFR and related kinase families
Thrombosis and coagulation — thrombin, factor Xa, FXIa, plasma kallikrein	GPCR programs (not yet characterised)
Antibacterial proteases — ClpP, Lon, SpsB, serine beta-lactamase	Nuclear receptor programs (not yet characterised)
Other proteases — renin, plasmepsin, cruzipain, cathepsins	—

Reproducibility

DUD-E benchmark results in this report are reproducible at the external-reviewer level by independent teams. Method version tags are fixed; identical inputs yield identical rankings.

Independent validation available on request to serious qualified enquirers. Reviewers receive a reproducibility pack sufficient to regenerate the numbers in this report on their own hardware, together with a technical briefing on architectural properties relevant to evaluation.

Need the full technical report?

Request the reproducibility pack and per-target breakdown. We share it with serious qualified enquirers along with trial credentials.

This document is for public distribution. Revisions: v1.1 will be issued alongside the v1.1 product release with updated kinase benchmarks.