About - glarue/intronIC GitHub Wiki

About

intronIC was created by Graham Larue during his PhD at UC Merced to provide a customizable, open-source method for identifying minor (U12-type) spliceosomal introns from annotated intron sequences.

Background

Minor introns usually represent ~0.5% (at most) of a given genome's introns, and contain distinct splicing motifs which make them amenable to bioinformatic identification. Despite their rarity, U12-type introns are:

  • Functionally important: Genes containing U12-type introns are often involved in essential cellular processes
  • Evolutionarily conserved: Present in most eukaryotic lineages, with documented losses in nematodes, some fungi, and certain protists

Motivation

Earlier minor intron resources (U12DB, SpliceRack, ERISdb, etc.), while important contributions to the field, are static by design. These databases:

  • Are not updated with new genome releases
  • Are based on older genome annotations
  • Are limited to pre-selected species
  • Use heuristic classification criteria

intronIC addresses these limitations by:

  • Working with any genome + annotation
  • Using well-established SVM classification with a pretrained model
  • Producing interpretable probability scores
  • Supporting custom model training for specialized use cases
  • Providing per-intron metadata for downstream analysis
  • Updated as the algorithm develops

Classification approach

intronIC relegates the "how U12-like" question to a support vector machine (SVM) ensemble. The bundled v3 default is a 42-model RBF-kernel SVM ensemble trained on a multi-species corpus (see Training data and PWMs). It scores each intron's background-corrected raw motif log-odds (5′SS, branch point, 3′SS) plus branch-point geometry into P_motif, a species-agnostic, Platt-calibrated U12-type motif probability. A separate output-level species adjudicator then makes the per-species population call (motif_category) and writes the final adjusted_score calling column. The pipeline:

  • Learns a non-linear decision boundary from curated reference data
  • Produces a species-agnostic probability (P_motif) that a user can threshold post-hoc at any confidence level
  • Uses balanced class weights to handle extreme class imbalance (~0.5% U12-type)
  • Operates directly on the raw motif log-odds, with no per-species z-normalization or recalibration

This avoids arbitrary score thresholds and provides probabilistic classifications that researchers can interpret based on their specific needs (e.g., high-confidence predictions for experimental validation vs. comprehensive catalogs).

Metadata and bookkeeping

intronIC records metadata during intron collection, including:

  • Parent gene/transcript identifiers
  • Ordinal intron index within transcript
  • Intron phase (position relative to codon boundary)
  • Fractional position within the coding sequence
  • Terminal dinucleotides and motif sequences

This information, which is otherwise non-trivial to acquire, enables downstream analyses of U12-type intron distribution, positional biases, and evolutionary patterns.

Citation

If you use intronIC in your research, please cite:

Moyer DC, Larue GE, Hershberger CE, Roy SW, Padgett RA. (2020) Comprehensive database and evolutionary dynamics of U12-type introns. Nucleic Acids Research 48(13):7066–7078. doi:10.1093/nar/gkaa464

License

intronIC is released under the GNU General Public License v3.0.

Related Resources