The role of small in-frame insertions/deletions in inherited eye disorders and how structural modelling can help estimate their pathogenicity

Table 1 Small in-frame insertions/deletions for which reliable structural models could be generated

Gene	Sequence change	Protein change	Template used for integrative structural modeling	Structural modeling prediction: does this change disrupt protein structure/ function?	Clinical report: does this change account for the clinical presentation?	Is there agreement between in silico tools for this change?
FSCN2	c.1071_1073del	p.(Lys357del)	human FSCN1 (pdb 1DFC)	unlikely	unlikely	yes [D/D/D]
RP2	c.260_268del	p.(Thr87_Cys89del)	human RP2 (pdb 2BX6)	probably	possibly	yes [D/D/D]
RPE65	c.1443_1445del	p.(Glu481del)	cow RPE65 (pdb 3FSN)	unclear	probably	yes [D/D/D]
BFSP2	c.697_699del	p.(Glu233del)	human vimentin (pdb 3UF1)	probably	probably	yes [D/D/D]
CRYBA1	c.272_274del	p.(Gly91del)	human CRYBA4 (pdb 3LWK)	probably	probably	yes [D/D/D]
CRYBA4	c.136_156del	p.(Ser46_Gly52del)	human CRYBA4 (pdb 3LWK)	probably	probably	yes [D/D/D]
CRYGC	c.61_63del	p.(Thr21del)	human CRYGB (pdb 2JDF)	unclear	probably	yes [D/D/D]
PITX2	c.429_431del	p.(Arg144del)	human PITX2 (pdb 2LKX)	probably	probably	no [N/D/D]

Assuming the clinical report is the standard and after removing the case where the variant possibly accounted for the clinical presentation (RP2 p. (Thr87_Cys89del)), the test accuracy was found to be 0.86 for structural modeling, SIFT-indel and PROVEAN, and 0.71 for DDIG-in. SIFT-indel and PROVEAN had the highest sensitivity (1.00) while structural modeling had the highest specificity (0.75)
[D/D/D] suggests that an in-frame indel was predicted to be disease-associated by DDIG-in, damaging by SIFT-indel and deleterious by PROVEAN; [N/D/D] suggests that it was predicted to be neutral by DDG-in, damaging by SIFT-indel and deleterious by PROVEAN
For more details on transcripts, RCSB PDB entries, and in silico analysis please see text and Additional file 1: Table S2

ISSN: 1750-1172