Struct Annotation 

pub struct Annotation {
    pub chr_exon_start_map: HashMap<String, String>,
    pub chr_exon_end_map: HashMap<String, String>,
    pub chr_exon_start_index: HashMap<String, HashMap<i32, (String, char)>>,
    pub chr_exon_end_index: HashMap<String, HashMap<i32, (String, char)>>,
    pub gene_exon_map: HashMap<String, Vec<(i32, i32)>>,
    pub gene_transcript_exon_map: HashMap<String, Vec<Vec<(i32, i32)>>>,
    pub transcript_splice_map: HashSet<String>,
    pub transcript_splice_index: HashMap<String, HashMap<char, HashSet<(i32, i32)>>>,
    pub chr_gene_map: HashMap<String, Vec<GeneSpan>>,
}

Parsed GTF-derived lookup tables used during BSJ validation and final labeling.

The current representation is intentionally string-keyed because Scan1 hot paths already build Java-shaped "chr\tpos" keys; keeping that contract avoids repeated format conversions during parity-sensitive checks.

Fields

chr_exon_start_map: HashMap<String, String>

Maps from genomic coordinate string (“Chr\tStart”) to its exon metadata (“GeneID\tStrand”).

chr_exon_end_map: HashMap<String, String>

Maps from genomic coordinate string (“Chr\tEnd”) to its exon metadata (“GeneID\tStrand”).

chr_exon_start_index: HashMap<String, HashMap<i32, (String, char)>>

Numeric exon-start lookup for post-Summary CIRI-AS boundary correction.

Scan1/Scan2 keep using the Java-shaped string maps above for parity. Segments reconstruction tests thousands of nearby splice-site candidates, so it uses this side index to avoid formatting "chr\tpos" strings in the hot loop.

chr_exon_end_index: HashMap<String, HashMap<i32, (String, char)>>

Numeric exon-end lookup for post-Summary CIRI-AS boundary correction.

gene_exon_map: HashMap<String, Vec<(i32, i32)>>

Maps from Gene ID to a vector of its exons (as start and end pairs).

gene_transcript_exon_map: HashMap<String, Vec<Vec<(i32, i32)>>>

Transcript-consistent exon chains grouped by gene ID.

gene_exon_map keeps the historical gene-level union used by Summary fallback labels. Full-length sidecar reconstruction needs a stricter projection source, because mixing all transcript exons from one gene can create impossible exon chains inside unsupported blocks.

transcript_splice_map: HashSet<String>

Transcript-consistent introns keyed as chr\tleft_exon_end\tright_exon_start\tstrand.

This is intentionally a side lookup rather than a replacement for the Java-shaped exon-boundary maps above: CIRI3 parity paths still need the boundary-only behavior, while CIRI-AS read-level segments can use adjacent exon pairs from the same transcript to resolve ambiguous splice offsets.

transcript_splice_index: HashMap<String, HashMap<char, HashSet<(i32, i32)>>>

Numeric transcript-consistent splice-pair index for CIRI-AS hot loops.

chr_gene_map: HashMap<String, Vec<GeneSpan>>

Groups genes by chromosome with their outer exon span for Summary’s exon/intron/intergenic fallback labeling.

Implementations

impl Annotation

pub fn new() -> Self

Initializes an empty annotation store.

pub fn read_gtf(&mut self, annotation_file: &str) -> Result<()>

Reads a GTF file and populates the exon-boundary lookup tables.

Only exon features are retained because the rest of the pipeline only needs exon boundary membership and gene IDs for parity with CIRI3.

Arguments

• annotation_file - Path to the input GTF file.

pub fn find_exon_match( &self, chr: &str, tmp_site1: i32, tmp_site2: i32, adjt_bp: i32, ) -> Option<(i32, String, String, String)>

Checks whether a shifted junction pair matches known exon boundaries.

This helper is mainly useful for tests and experiments; Scan1’s hot path now performs the same lookup inline to stay closer to Java branch order.

Arguments

• chr - Chromosome name.
• tmp_site1 - Predicted upstream junction coordinate.
• tmp_site2 - Predicted downstream junction coordinate.
• adjt_bp - Range of shift offsets to check.

Returns

Some(shift, strand, donor_placeholder, acceptor_placeholder) if a match is found.