Examples of transcript-confirmed intron predictions. (A) A predicted intron is located in the 5′ UTR of the protein-coding gene CG14614, whose current 5′ UTR annotation consists of only 2 nt. (B) Example of a predicted intron that belongs to a transcript overlapping an intron of dally in the antisense direction. (C) Example of a predicted intron that belongs to a potentially tissue-specific noncoding RNA, as 13 of the 14 supporting ESTs originate from a salivary gland library (ESG01). (D) A predicted intron that overlaps a noncoding FlyBase transcript (pncr009:3L) that has no intron annotation. pncr009:3L was found to be a structured precursor for small interfering RNAs (Okamura et al. 2008). (E) Example of a “cluster” of three introns within ∼400 nt. All three introns are predicted with a probability of >0.999 and belong to a potentially coding gene (BLASTX hits in several Drosophila species). Examples B–E illustrate that our approach finds introns that are located in regions of low sequence conservation, indicated by low phastCons conservation scores up- and downstream of the intron. Modified UCSC genome browser (Karolchik et al. 2008) screenshots were used to make this figure.