Description

The NMDetective tracks display genome-wide predictions of nonsense-mediated mRNA decay (NMD) efficiency from Lindeboom et al. 2016. NMDetective scores predict whether a premature termination codon (PTC) at a given position will trigger NMD and mRNA degradation, or whether the transcript will escape NMD and potentially produce a truncated protein.

Scores range from approximately −1 to +1. Positive values indicate that a PTC at that position is predicted to trigger NMD (the mRNA is degraded). Negative values indicate that the PTC is predicted to escape NMD (the truncated mRNA may be translated into an aberrant protein). Values near zero indicate intermediate or uncertain NMD efficiency.

Subtracks

Track	Description
NMDetective-A	Random forest model predicting NMD efficiency for all possible PTCs introduced by single-nucleotide variants. Explains ~71% of systematic variance in NMD efficiency.
NMDetective-B	Simplified decision tree model for all possible PTCs. Slightly lower accuracy (~68% variance explained) but more interpretable, making it suitable for clinical applications.
NMDetective-A PTC	Random forest model predicting NMD efficiency specifically for the first out-of-frame PTC introduced by frameshifting indel mutations.
NMDetective-B PTC	Decision tree model for the first out-of-frame PTC from frameshifting indels.

Display Conventions and Configuration

Each subtrack is displayed as a signal (bigWig) track. By default, the vertical axis ranges from −1 to +1. Regions with positive values (predicted NMD-triggering) are shown above the baseline; regions with negative values (predicted NMD escape) are shown below.

Blue tracks (NMDetective-A and -B): predictions for all possible PTCs from single-nucleotide nonsense variants.
Green tracks (NMDetective-A PTC and -B PTC): predictions for the first out-of-frame PTC from frameshifting indels.

Methods

The NMDetective models were trained on somatic nonsense mutation data from 9,769 cancer patients and validated with frameshift mutations and germline variants (Lindeboom et al. 2019). The models incorporate the following features to predict NMD efficiency:

Whether the PTC falls in the last exon
Distance to the last 50 nt of the penultimate exon (the EJC-based “50 bp rule”)
Distance from the coding start (start-proximal NMD insensitivity)
Exon length
mRNA half-life
Distance to the downstream exon-junction complex
Distance to the wild-type stop codon

NMDetective-A (random forest regression) captures non-linear interactions among these features and achieves the highest predictive accuracy. NMDetective-B (decision tree) applies a simpler rule-based classification that is more transparent, with a modest reduction in accuracy.

The predictions were generated for every possible PTC-introducing single-nucleotide variant and for the first out-of-frame PTC from every possible single-nucleotide frameshifting indel across all human protein-coding transcripts. The original bedGraph custom track files were downloaded from the NMDetective Figshare page resource and converted to bigWig format at UCSC.

Data Access

The data underlying these tracks can be explored interactively with the Table Browser or the Data Integrator. For automated analysis, the data may be queried from our REST API. Please refer to our mailing list archives for questions, or our Data Access FAQ for more information.

Credits

Thanks to Rik Lindeboom for providing custom tracks and the original NMDetective data on Figshare.

References

Lindeboom RG, Supek F, Lehner B. The rules and impact of nonsense-mediated mRNA decay in human cancers. Nat Genet. 2016 Oct;48(10):1112-8. PMID: 27618451; PMC: PMC5045715

Lindeboom RGH, Vermeulen M, Lehner B, Supek F. The impact of nonsense-mediated mRNA decay on genetic disease, gene editing and cancer immunotherapy. Nat Genet. 2019 Nov;51(11):1645-1651. PMID: 31659324; PMC: PMC6858879