GnnDebugger: GNN based error correction in De Bruijn Graphs
Marijo Šimunović, Lovro Vrček, Mile Šikić, Anton Bankevich · 2026
Read GnnDebugger: GNN based error correction in De Bruijn Graphs by Marijo Šimunović, Lovro Vrček, Mile Šikić, Anton Bankevich (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13651195.txt.
Description
Marijo Šimunović and three colleagues, publishing in 2026, teach a neural network to proofread genome assembly graphs. Assemblers stitch DNA reads into graphs where errors create wrong turns. Correcting them by handcrafted rules struggles most in two-copy genomes at modest coverage, exactly where sequencing thousands of species gets hard.
Their program, GnnDebugger, uses a graph neural network, a model that learns patterns across connected nodes, to classify each graph edge as genuine or erroneous. The classifier proved especially reliable for diploid genomes at coverage of thirty-five-fold or lower. Feeding its verdicts into a standard read-correction and assembly pipeline produced more accurate assemblies than the heuristics alone. The authors argue that learned correction can replace rule-based cleanup in difficult scenarios and adapt more readily to newly sequenced species, helping scale complete genome projects beyond model organisms.
Catalog record
| CREATOR | Marijo Šimunović, Lovro Vrček, Mile Šikić, Anton Bankevich |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53463 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |