ingap-cdg

ingap-cdg constructs full-length, non-redundant coding sequences (CDSs) directly from unassembled transcriptome reads to enable accurate gene prediction without closely related reference genomes.

Key Features:

• Codon-Based De Bruijn Graph: Employs a codon-based de Bruijn graph to assemble coding sequences from fragmented transcriptomic data.
• Machine Learning Integration: Uses machine learning algorithms to filter false positives and distinguish true coding sequences from sequencing artifacts.
• Robustness to Sequencing Errors: Maintains prediction accuracy despite sequencing errors and variability in read lengths.
• Increased Predicted CDS Length: Produces longer predicted coding sequences compared to other existing methods, improving completeness of gene representations.

Scientific Applications:

• Reference-limited transcriptome analysis: Enables CDS prediction in studies lacking closely related reference genomes, including non-model organisms and newly discovered species.
• Gene structure and function inference from raw reads: Facilitates reconstruction of coding sequences directly from raw transcriptome reads to support downstream gene structure and functional analyses.

Methodology:

Assembly simplification using a codon-based de Bruijn graph to construct coding sequences from fragmented transcriptomic data, and false-positive filtering using machine learning algorithms to distinguish true CDSs from artifacts.