Figure 4. Latent class analysis of the unannotated translated open reading frames (utORFs) reveals differences between classes.
(A) Class 1 is notably distinct for strong bias toward intergenic and antisense locations at the expense of sense locations. Class 2 is notable for being relatively unbiased and for being the only class with appreciable members in a combination of locations. Class 5 is strongly skewed toward antisense and sense locations. (B) Class 1 is almost exclusively of intermediate length. Class 4 has the greatest length bias, followed by class 2. Short: fewer than 20 residues; medium: from 20 to 49 residues; long: 50 or more residues. (C) Class 1 is notably distinct from the others for minimal transcription. Low tissue and developmental specificity may be an effect of minimal transcription. Class 2 is remarkable for being entirely monophyletic. Class 5 has slightly lower tissue and developmental specificities than classes 2–4. significant TPM: maximum per-sample transcripts per million (TPM) > 0.1; tissue specificity: tissue specificity > 0.8; developmental specificity: developmental specificity > 0.8. (D) Class 2 is by far the youngest. Class 3 tends to be of an intermediate age, with inferred emergence at around the latent class analysis (LCA) of the melanogaster subgroup (D. rho). In contrast, class 4 emergence is distributed throughout the LCA of the melanogaster subgroup and the Drosophila genus. (E) Class 2 is notable for overall low conservation. Class 3 is remarkably even less conserved. Classes 4 and 5 are distinguished through differences in intermediate vs. significant conservation. nonconserved: phastCons score < 0.2; ambiguous: phastCons score ≥ 0.2 and < 0.8; conserved: phastCons score ≥ 0.8.
Figure 4—figure supplement 1. Posterior probabilities per unannotated translated open reading frame (utORF) of class membership inferred from latent class analysis for all utORFs.
Figure 4—figure supplement 2. Latent class analysis of unannotated translated open reading frames (utORFs) with canonical start sites reveals differences between classes.
Figure 4—figure supplement 3. Posterior probabilities per unannotated translated open reading frame (utORF) of class membership inferred from latent class analysis for utORFs with canonical start sites.
