Evolutionary relationships between prodomains in the TGF-β family have gone unanalyzed due to a perceived lack of conservation. We developed a novel approach, identified these relationships, and suggest hypotheses for new regulatory mechanisms in TGF-β signaling. First, a quantitative analysis placed each family member from flies, mice, and nematodes into the Activin, BMP, or TGF-β subfamily. Second, we defined the prodomain and ligand via the consensus cleavage site. Third, we generated alignments and trees from the prodomain, ligand, and full-length sequences independently for each subfamily. Prodomain alignments revealed that six structural features of 17 are well conserved: three in the straitjacket and three in the arm. Alignments also revealed unexpected cysteine conservation in the “LTBP-Association region” upstream of the straitjacket and in β8 of the bowtie in 14 proteins from all three subfamilies. In prodomain trees, eight clusters across all three subfamilies were present that were not seen in the ligand or full-length trees, suggesting prodomain-mediated cross-subfamily heterodimerization. Consistency between cysteine conservation and prodomain clustering provides support for heterodimerization predictions. Overall, our analysis suggests that cross-subfamily interactions are more common than currently appreciated and our predictions generate numerous testable hypotheses about TGF-β function and evolution.
Reduced Annotation Based on Deep Active Learning for Arabic Text Detection in Natural Scene Images