Whole-Exome Sequencing in a Consanguinity-Enriched South Indian Retinitis Pigmentosa Cohort: Diagnostic Yield and Molecular Spectrum Authors: Sujay Jaju, Jayashree Sivagnanam, Parvathy Menon, Elakkiya Nandini G R, Padmapriya Shanmugam, Anju Jose, Rodney Morris, Kim Ramasamy, Narendran Venkatapathy, Jaishree Pandian, Siddharth Narendran Published: 2026 ABSTRACT PurposeTo determine the molecular diagnostic yield, variant spectrum, inheritance architecture, and influence of consanguinity on whole-exome sequencing outcomes in a South Indian retinitis pigmentosa (RP) cohort.DesignProspective, registry-based cohort study.SubjectsA total of 113 affected participants were enrolled through the Aravind Registry for Inherited Diseases of the Eye, including 109 unrelated probands and 4 affected relatives from already represented families. Primary analyses were restricted to the 109 unrelated probands.MethodsWhole-exome sequencing was performed using a clinical exome workflow. Variants were interpreted using American College of Medical Genetics and Genomics/Association for Molecular Pathology criteria and cases were categorized as solved, possibly solved, inconclusive, or unsolved using prespecified inheritance-aware rules.Main Outcome MeasuresMolecular diagnostic yield, distribution of implicated genes and variant classes, inheritance architecture, and diagnostic yield stratified by consanguinity status.ResultsAmong the 109 unrelated probands, mean age at testing was 39.3 ± 14.1 years and 58.7% were male. Whole-exome sequencing identified 186 distinct rare variants across 92 inherited retinal disease genes, including 26 pathogenic and 33 likely pathogenic variants. A molecular diagnosis was established in 50 of 109 probands (45.9%), including 42 solved and 8 possibly solved cases; 45 (41.3%) were inconclusive and 14 (12.8%) remained unsolved, including 4 (3.7%) in whom no candidate variant was identified. EYS, USH2A, and ADGRV1 were the most frequently implicated genes. Autosomal recessive (AR) disease predominated (44/50, 88.0%). Consanguineous AR cases were exclusively homozygous (17/17); notably, 68.0% of nonconsanguineous AR cases were also homozygous (P = 0.013). Diagnostic yield was higher in consanguineous probands (51.4% vs. 41.7%), without reaching significance. Recurrent alleles included an established South Asian founder variant (MFSD8 c.1361T>C) and candidate founder alleles in EYS (c.4321C>T) and ADGRV1 (c.14329C>T).ConclusionsWhole-exome sequencing established a molecular diagnosis in nearly half of this South Indian RP cohort and revealed a predominantly recessive, homozygosity-enriched architecture shaped by consanguinity. These findings define a region-specific variant landscape to support clinical interpretation, genetic counseling, and future trial enrollment in this underrepresented population.Financial Disclosure(s)The authors have no proprietary or commercial interest in any materials discussed in this article. Retinitis pigmentosa (RP) is the most prevalent inherited retinal disease (IRD) and a leading cause of hereditary blindness, affecting >1.5 million people worldwide.1,2 It is associated with substantial reductions in health-related quality of life and a high lifetime economic burden, driven by increased health care utilization, reduced employment, and lost productivity.3, 4, 5, 6, 7, 8 These clinical and socioeconomic burdens, together with RP’s status as the most common IRD, have made it a major focus of translational research and of emerging gene- and cell-based therapies. However, the current IRD therapeutic-trial landscape remains concentrated in high-income countries, with comparatively limited participation from low- and middle-income settings.9, 10, 11 Population-based surveys in South India estimate one of the highest reported prevalences of RP globally (∼1 in 372), nearly an order of magnitude higher than Western populations (∼1 in 3000–5000).2,11,12 This disparity is attributed in part to consanguineous marriage practices, which are reported in ∼23% of South Indian families and represent a major demographic driver of autosomal recessive (AR) retinal disease.13,14 The rapid expansion of gene-specific and variant-specific IRD therapies has made precise molecular characterization an increasingly critical prerequisite for trial eligibility, genetic counseling, and therapy access planning.9 Despite this high burden, South Indian patients remain markedly underrepresented in global genomic reference datasets and IRD clinical trials, and the existing Indian IRD literature derives predominantly from retrospective or gene-targeted studies rather than prospective whole-exome sequencing (WES) cohorts.15, 16, 17 In the absence of region-specific variant data, patients risk both misclassification of clinically significant variants and systematic exclusion from emerging gene therapy programs.9,17 To begin addressing these gaps, the Aravind Registry for Inherited Diseases of the Eye (ARIDE) was established as a prospective, hospital-based program to systematically capture clinical and genomic information on IRDs across South India. Here we report the following WES findings from a consanguinity-enriched RP cohort to: (1) estimate molecular diagnostic yield; (2) define the regional causal gene and variant spectrum; and (3) characterize the influence of consanguinity on inheritance architecture and diagnostic outcome. Methods Study Design and Cohort This prospective, registry-based study enrolled 113 affected participants with RP through ARIDE at Aravind Eye Hospital, Coimbatore, Tamil Nadu, India, including 109 unrelated probands and 4 affected relatives from already represented families. Patient recruitment for this analysis began on September 2, 2024, and continued through March 31, 2025. The registry is registered with the Clinical Trials Registry of India (CTRI/2024/07/071686). Consecutive unrelated probands enrolled in ARIDE with a clinical diagnosis of nonsyndromic RP, syndromic RP, atypical RP (including RP sine pigmento), or rod–cone dystrophy were eligible for the primary analysis, irrespective of their ability to pay for genetic testing; WES was funded through the ARIDE registry program for participants who required financial support, while others self-funded testing, and ability to pay was neither an eligibility criterion nor a determinant of enrollment. Four additional affected relatives from already represented families were included for family-based interpretation and descriptive purposes. Primary molecular diagnostic yield, mutation-spectrum, inheritance, and consanguinity analyses were restricted to the 109 unrelated probands. The cohort comprised both adult and pediatric participants; written informed consent was obtained from adult patients and from parents or legal guardians of minors. The study was approved by the Institutional Human Ethics Committee of the PSG Institute of Medical Sciences and Research, Coimbatore (approval reference PSG/IHEC/2024/Appr/FB/043), and adhered to the tenets of the Declaration of Helsinki. Participants with suspected IRD were identified following a detailed family history assessment and underwent a standardized ophthalmic examination comprising best-corrected visual acuity testing, slit lamp biomicroscopy, tonometry, Ishihara color vision testing, widefield fundus photography (Clarus, Carl Zeiss Meditec, Inc), spectral-domain OCT (Spectralis HRA + OCT, Heidelberg Engineering, Inc), and full-field electroretinography (Diagnosys LLC) when clinically indicated. WES and Variant Interpretation Peripheral blood (2 mL, ethylenediaminetetraacetic acid) was collected from each proband and participating family members, and genomic DNA was extracted using standard protocols. Whole-exome sequencing was performed at a commercial clinical genomics laboratory (Nucleome Informatics) using the Twist Exome 2.0 Capture Kit (Twist Bioscience; target size approximately 36 Mb) and sequencing on an Illumina NovaSeq 6000 platform with 2 × 150 base pair paired-end reads. Primary bioinformatics processing followed the clinical laboratory workflow. Reads were processed using the Illumina Dynamic Read Analysis for Genomics Bio-IT Platform and aligned to the Genome Reference Consortium human genome build 38/human genome 38 reference genome. Across samples, read alignment exceeded 99%, the proportion of bases at Phred quality greater than or equal to Q30 was approximately 90%–92%, mean on-target depth was approximately 50–80×, and >94% of target bases were covered at ≥20×. Sequence variants (single-nucleotide variants and small insertion or deletions) were called using a validated pipeline including HaplotypeCaller. Copy-number and exon-level structural variants were identified from exome read-depth using Dynamic Read Analysis for Genomics copy-number analysis and classified according to the American College of Medical Genetics and Genomics (ACMG)/Clinical Genome Resource technical standards for constitutional copy-number variants, separately from the aforementioned sequence-variant criteria.18 Read-depth copy-number calls were not orthogonally confirmed by multiplex ligation-dependent probe amplification or quantitative polymerase chain reaction, and cases resting on an unconfirmed copy-number variant were classified as possibly solved. Candidate variants were filtered against population frequency databases (the Genome Aggregation Database, 1000 Genomes, the Single Nucleotide Polymorphism Database) and annotated using Human Genome Variation Society nomenclature via Ensembl Variant Effect Predictor; in silico pathogenicity support used Sorting Intolerant From Tolerant, Polymorphism Phenotyping v2, and MutationTaster2.19 Variants were interpreted according to the ACMG/Association for Molecular Pathology (AMP) criteria. Variants with low quality (Phred quality <30 or depth <10×) were excluded, and all reported variants were manually reviewed in Integrative Genomics Viewer to exclude sequencing or alignment artifacts.20 Identified variants underwent the ACMG/AMP interpretation.21 A case was considered “solved” when: (1) The proband harbored a pathogenic (P) or likely pathogenic (LP) variant (ACMG/AMP classification) in a gene known to cause autosomal dominant IRD or, in male patients, in an X-linked IRD gene; or (2) The proband harbored 2 pathogenic or LP variants in a gene known to cause AR IRD, consistent with the expected inheritance pattern (biallelic state inferred from genotype when parental data were unavailable). A case was considered “partially solved/possibly solved” when the proband carried 2 variants in a recessive IRD gene, one classified as pathogenic or LP and the other as a variant of uncertain significance (VUS). A case was classified as “inconclusive” when only VUS were identified in genes relevant to the IRD phenotype. Cases without any credible candidate variants in known IRD genes, or those with a single heterozygous pathogenic/LP variant in a recessive gene, were considered “unsolved.” Statistical Analysis Continuous variables were summarized as mean ± standard deviation or median (range), and categorical variables as counts and percentages. Primary diagnostic-yield, mutation-spectrum, inheritance, and consanguinity analyses were restricted to the 109 unrelated probands. Differences in diagnostic yield between consanguineous and nonconsanguineous probands were assessed using Fisher exact test; a 2-sided P value <0.05 was considered statistically significant. All analyses were performed in R version 4.3 (R Foundation for Statistical Computing). Results Cohort Characteristics Primary analyses were performed on 109 unrelated probands from the 113 affected participants enrolled, which also included 4 affected relatives from already represented families. The mean age at genetic testing was 39.3 ± 14.1 years (median 38; range, 15–72 years), and 64 probands (58.7%) were male. The mean best-corrected visual acuity was 0.65 ± 0.67 logarithm of the minimum angle of resolution in the right eye and 0.61 ± 0.61 logarithm of the minimum angle of resolution in the left eye (median 0.40 and 0.32, respectively; range 0.00–2.70 in both eyes). Clinically, 91 probands (83.5%) had nonsyndromic RP, 10 (9.2%) had syndromic RP, 7 (6.4%) had atypical RP, and 1 proband (0.9%) was classified as rod–cone dystrophy. Parental consanguinity was reported in 35 probands (32.1%), absent in 72 (66.1%), and unavailable in 2 (1.8%). Diagnostic Yield Among the 109 unrelated probands, WES identified 186 distinct rare variants in 92 genes known or suspected to be associated with IRD. Of these, 26 (14.0%) were classified as pathogenic and 33 (17.7%) as LP according to ACMG/AMP criteria; the remaining 127 (68.3%) were VUS. Applying prespecified inheritance-aware adjudication rules, 42 (38.5%) probands were classified as solved and 8 (7.3%) as possibly solved, yielding an overall molecular diagnosis in 50 of 109 probands (45.9%). Forty-five probands (41.3%) remained inconclusive because only phenotype-relevant VUS were identified, and 14 (12.8%) were unsolved, including 4 in whom no rare variant was detected in known IRD genes. The distribution of adjudication outcomes across the full cohort and stratified by consanguinity status is shown in Figure 1A and Table 1.Figure 1Molecular diagnostic yield and causal gene spectrum in the ARIDE South Indian retinitis pigmentosa cohort. (A) Proportional stacked horizontal bars showing case-adjudication outcomes among 109 unrelated probands overall and after stratification by reported parental consanguinity. The upper bar shows all probands; the lower bars show consanguineous (n = 35) and nonconsanguineous (n = 72) probands. Two probands with undetermined consanguinity are excluded from the stratified display. Outcome categories are solved, possibly solved, inconclusive, unsolved, and no variant detected. The overall molecular diagnostic yield (solved + possibly solved) was 45.9% (50/109), including 51.4% (18/35) in consanguineous probands and 41.7% (30/72) in nonconsanguineous probands. (B) Horizontal bar chart showing the number of diagnosed probands per causal gene among genes implicated in ≥2 of the 50 diagnosed probands. EYS and USH2A were the most frequently implicated genes (n = 7), followed by ADGRV1 (n = 4) and MFSD8 (n = 3); BBS10, MYO7A, NR2E3, PDE6A, PROM1, RHO, and RPGR were each implicated in 2 probands. Genes implicated in only 1 diagnosed proband are not shown. ARIDE = Aravind Registry for Inherited Diseases of the Eye.Table 1Cohort Characteristics, Diagnostic Outcomes, Inheritance Architecture, and Whole-Exome Sequencing Findings in 109 Unrelated Probands with Retinitis Pigmentosa (Aravind Registry for Inherited Diseases of the Eye)CharacteristicAll Probands (n = 109)Consanguineous (n = 35)Nonconsanguineous (n = 72)A. Demographics Age, mean ± SD, years39.3 ± 14.134.7 ± 13.541.6 ± 14.1 Age, median (range), years38 (15–72)31 (15–66)42 (16–72) Male sex, n (%)64 (58.7)22 (62.9)40 (55.6)B. Case adjudication Solved, n (%)42 (38.5)18 (51.4)22 (30.6) Possibly solved,⁠ n (%)8 (7.3)0 (0.0)8 (11.1) Total diagnosed,⁠ n (%)50 (45.9)18 (51.4)30 (41.7) Inconclusive, n (%)45 (41.3)14 (40.0)31 (43.1) Unsolved, n (%)14 (12.8)3 (8.6)11 (15.3) No variant detected, n (%)4 (3.7)1 (2.9)3 (4.2) Diagnostic yield OR (95% CI)∗—1.48 (0.66–3.34)P = 0.41C. Inheritance in diagnosed cases [denominator: 50 diagnosed probands]† Autosomal recessive (AR), n (%)44 (88.0)17 (94.4)25 (83.3) Homozygous biallelic, n (% of AR)36 (81.8)17 (100.0)17 (68.0) Likely compound heterozygous, n (% of AR)8 (18.2)0 (0.0)8 (32.0) Autosomal dominant (AD), n (%)4 (8.0)1 (5.6)3 (10.0) X-linked recessive (XL), n (%)2 (4.0)0 (0.0)2 (6.7) Hom vs. comp het in AR: Fisher exact⁠∗—17/17 hom (100%)17/25 hom (68.0%); P = 0.013D. Whole-exome sequencing findings‡ [denominator: 186 distinct variants] Distinct rare variants (genes implicated), n186 (92)63 (43)127 (69) Pathogenic (P), n (%)26 (14.0)4 (6.3)23 (18.1) Likely pathogenic (LP), n (%)33 (17.7)18 (28.6)16 (12.6) P + LP combined, n (%)59 (31.7)22 (34.9)39 (30.7) Variant of uncertain significance, n (%)127 (68.3)41 (65.1)88 (69.3)cHet = likely compound heterozygous; CI = confidence interval; Hom/hom = homozygous; OR = odds ratio; SD = standard deviation.∗Odds ratio (95% CI) calculated by Fisher exact test.†Section C denominators: the 2 undetermined-consanguinity probands are excluded from Section C stratified columns, consistent with Sections A–B.‡Six variants occurred in both consanguineous and nonconsanguineous probands (EYS c.4321C>T, MFSD8 c.1361T>C, PDE6A c.1151dupA, ABCA4 c.6694G>A, PROM1 c.1946C>T, SRD5A3 c.233A>G) and are counted in both stratified columns; 2 variants occurred only in the 2 probands of undetermined consanguinity. Stratified columns therefore exceed 186. Molecular diagnostic yield and causal gene spectrum in the ARIDE South Indian retinitis pigmentosa cohort. (A) Proportional stacked horizontal bars showing case-adjudication outcomes among 109 unrelated probands overall and after stratification by reported parental consanguinity. The upper bar shows all probands; the lower bars show consanguineous (n = 35) and nonconsanguineous (n = 72) probands. Two probands with undetermined consanguinity are excluded from the stratified display. Outcome categories are solved, possibly solved, inconclusive, unsolved, and no variant detected. The overall molecular diagnostic yield (solved + possibly solved) was 45.9% (50/109), including 51.4% (18/35) in consanguineous probands and 41.7% (30/72) in nonconsanguineous probands. (B) Horizontal bar chart showing the number of diagnosed probands per causal gene among genes implicated in ≥2 of the 50 diagnosed probands. EYS and USH2A were the most frequently implicated genes (n = 7), followed by ADGRV1 (n = 4) and MFSD8 (n = 3); BBS10, MYO7A, NR2E3, PDE6A, PROM1, RHO, and RPGR were each implicated in 2 probands. Genes implicated in only 1 diagnosed proband are not shown. ARIDE = Aravind Registry for Inherited Diseases of the Eye. Table 1Cohort Characteristics, Diagnostic Outcomes, Inheritance Architecture, and Whole-Exome Sequencing Findings in 109 Unrelated Probands with Retinitis Pigmentosa (Aravind Registry for Inherited Diseases of the Eye)CharacteristicAll Probands (n = 109)Consanguineous (n = 35)Nonconsanguineous (n = 72)A. Demographics Age, mean ± SD, years39.3 ± 14.134.7 ± 13.541.6 ± 14.1 Age, median (range), years38 (15–72)31 (15–66)42 (16–72) Male sex, n (%)64 (58.7)22 (62.9)40 (55.6)B. Case adjudication Solved, n (%)42 (38.5)18 (51.4)22 (30.6) Possibly solved,⁠ n (%)8 (7.3)0 (0.0)8 (11.1) Total diagnosed,⁠ n (%)50 (45.9)18 (51.4)30 (41.7) Inconclusive, n (%)45 (41.3)14 (40.0)31 (43.1) Unsolved, n (%)14 (12.8)3 (8.6)11 (15.3) No variant detected, n (%)4 (3.7)1 (2.9)3 (4.2) Diagnostic yield OR (95% CI)∗—1.48 (0.66–3.34)P = 0.41C. Inheritance in diagnosed cases [denominator: 50 diagnosed probands]† Autosomal recessive (AR), n (%)44 (88.0)17 (94.4)25 (83.3) Homozygous biallelic, n (% of AR)36 (81.8)17 (100.0)17 (68.0) Likely compound heterozygous, n (% of AR)8 (18.2)0 (0.0)8 (32.0) Autosomal dominant (AD), n (%)4 (8.0)1 (5.6)3 (10.0) X-linked recessive (XL), n (%)2 (4.0)0 (0.0)2 (6.7) Hom vs. comp het in AR: Fisher exact⁠∗—17/17 hom (100%)17/25 hom (68.0%); P = 0.013D. Whole-exome sequencing findings‡ [denominator: 186 distinct variants] Distinct rare variants (genes implicated), n186 (92)63 (43)127 (69) Pathogenic (P), n (%)26 (14.0)4 (6.3)23 (18.1) Likely pathogenic (LP), n (%)33 (17.7)18 (28.6)16 (12.6) P + LP combined, n (%)59 (31.7)22 (34.9)39 (30.7) Variant of uncertain significance, n (%)127 (68.3)41 (65.1)88 (69.3)cHet = likely compound heterozygous; CI = confidence interval; Hom/hom = homozygous; OR = odds ratio; SD = standard deviation.∗Odds ratio (95% CI) calculated by Fisher exact test.†Section C denominators: the 2 undetermined-consanguinity probands are excluded from Section C stratified columns, consistent with Sections A–B.‡Six variants occurred in both consanguineous and nonconsanguineous probands (EYS c.4321C>T, MFSD8 c.1361T>C, PDE6A c.1151dupA, ABCA4 c.6694G>A, PROM1 c.1946C>T, SRD5A3 c.233A>G) and are counted in both stratified columns; 2 variants occurred only in the 2 probands of undetermined consanguinity. Stratified columns therefore exceed 186. Cohort Characteristics, Diagnostic Outcomes, Inheritance Architecture, and Whole-Exome Sequencing Findings in 109 Unrelated Probands with Retinitis Pigmentosa (Aravind Registry for Inherited Diseases of the Eye) cHet = likely compound heterozygous; CI = confidence interval; Hom/hom = homozygous; OR = odds ratio; SD = standard deviation. Odds ratio (95% CI) calculated by Fisher exact test. Section C denominators: the 2 undetermined-consanguinity probands are excluded from Section C stratified columns, consistent with Sections A–B. Six variants occurred in both consanguineous and nonconsanguineous probands (EYS c.4321C>T, MFSD8 c.1361T>C, PDE6A c.1151dupA, ABCA4 c.6694G>A, PROM1 c.1946C>T, SRD5A3 c.233A>G) and are counted in both stratified columns; 2 variants occurred only in the 2 probands of undetermined consanguinity. Stratified columns therefore exceed 186. Mutation Spectrum Among the 50 solved or possibly solved unrelated probands, 44 distinct putative causal variants were identified across 26 IRD genes (Fig 1B; Supplementary Table). EYS and USH2A were the most frequently implicated genes (7 cases; 14.0%), followed by ADGRV1 (4; 8.0%), and MFSD8 (3; 6.0%). Together, these 4 genes accounted for 21 of 50 diagnosed probands (42.0%). Seven additional genes, PROM1, NR2E3, RPGR, BBS10, MYO7A, RHO, and PDE6A, each contributed to 2 unrelated cases, while the remaining 15 genes were each implicated in a single case. Recurrent Variants Five pathogenic or LP variants recurred across unrelated probands and are detailed in Table 2 (Section A), alongside 5 recurrent VUS (Section B). EYS c.4321C>T (p.Gln1441Ter), a stop-gained variant, was identified in 4 unrelated probands; in 3 it was homozygous and disease-causing (all solved), and in one it was a heterozygous incidental carrier allele in a proband solved by biallelic ADGRV1 c.14329C>T. A second EYS variant, c.4045C>T (p.Arg1349Ter), also stop-gained and classified as pathogenic, was identified in 2 unrelated probands, homozygous in one (solved) and heterozygous in the other (a single allele only, unsolved). ADGRV1 c.14329C>T (p.Gln4777Ter), also a stop-gained variant, was identified in 3 independent probands, 2 homozygous (solved) and 1 compound heterozygous (possibly solved), and was classified as pathogenic in all instances. MFSD8 c.1361T>C (p.Met454Thr), a missense variant, was present in 3 unrelated probands, each homozygous and classified as pathogenic, all solved; 2 of the 3 carriers were from consanguineous families. PDE6A c.1151dupA (p.Asn384Lysfs∗24), a frameshift duplication, recurred in 2 unrelated probands, both homozygous and LP, both solved.Table 2Recurrent Variants in ≥2 Unrelated ProbandsGeneVariantsProtein Change (Consequence)ACMG ClassnZygosity in CarriersCons. Carriers, nA. Pathogenic (P) and likely pathogenic (LP) EYS∗c.4321C>Tp.Gln1441TerStop gainedP43 Hom1 Het1⁠+⁠1 undet. ADGRV1c.14329C>Tp.Gln4777TerStop gainedP32 Hom1 cHet0 EYSc.4045C>Tp.Arg1349TerStop gainedP21 Hom1 Het0 PDE6Ac.1151dupAp.Asn384Lysfs∗24Frameshift dup.LP22 Hom1 MFSD8c.1361T>Cp.Met454ThrMissenseP33 Hom2B. Variants of uncertain significance (VUS) PROM1c.1946C>Tp.Ser649LeuMissenseVUS22 Hom1 PROM1c.2231G>Cp.Arg744ThrMissenseVUS22 Het0 ABCA4c.6694G>Ap.Glu2232LysMissenseVUS21 Hom1 Het1 SRD5A3c.233A>Gp.His78ArgMissenseVUS22 Hom1 USH2Ac.13969T>Ap.Trp4657ArgMissenseVUS21 Hom1 cHet2ACMG = American College of Medical Genetics and Genomics; cHet = likely compound heterozygous; Cons. = consanguineous; Cons. carriers, n = number of variant carriers from consanguineous families; Het = heterozygous; Hom = homozygous; LP = likely pathogenic; n = number of probands with the variant; P = pathogenic; undet. = undetermined consanguinity status.∗EYS c.4321C>T was disease-causing (homozygous) in three unrelated probands. A fourth unrelated proband carried it heterozygously as an incidental allele (case solved by biallelic ADGRV1 c.14329C>T), and one further homozygous instance occurred in an excluded affected relative (Supplementary Table); neither is included in the causal count of three. Table 2Recurrent Variants in ≥2 Unrelated ProbandsGeneVariantsProtein Change (Consequence)ACMG ClassnZygosity in CarriersCons. Carriers, nA. Pathogenic (P) and likely pathogenic (LP) EYS∗c.4321C>Tp.Gln1441TerStop gainedP43 Hom1 Het1⁠+⁠1 undet. ADGRV1c.14329C>Tp.Gln4777TerStop gainedP32 Hom1 cHet0 EYSc.4045C>Tp.Arg1349TerStop gainedP21 Hom1 Het0 PDE6Ac.1151dupAp.Asn384Lysfs∗24Frameshift dup.LP22 Hom1 MFSD8c.1361T>Cp.Met454ThrMissenseP33 Hom2B. Variants of uncertain significance (VUS) PROM1c.1946C>Tp.Ser649LeuMissenseVUS22 Hom1 PROM1c.2231G>Cp.Arg744ThrMissenseVUS22 Het0 ABCA4c.6694G>Ap.Glu2232LysMissenseVUS21 Hom1 Het1 SRD5A3c.233A>Gp.His78ArgMissenseVUS22 Hom1 USH2Ac.13969T>Ap.Trp4657ArgMissenseVUS21 Hom1 cHet2ACMG = American College of Medical Genetics and Genomics; cHet = likely compound heterozygous; Cons. = consanguineous; Cons. carriers, n = number of variant carriers from consanguineous families; Het = heterozygous; Hom = homozygous; LP = likely pathogenic; n = number of probands with the variant; P = pathogenic; undet. = undetermined consanguinity status.∗EYS c.4321C>T was disease-causing (homozygous) in three unrelated probands. A fourth unrelated proband carried it heterozygously as an incidental allele (case solved by biallelic ADGRV1 c.14329C>T), and one further homozygous instance occurred in an excluded affected relative (Supplementary Table); neither is included in the causal count of three. Recurrent Variants in ≥2 Unrelated Probands ACMG = American College of Medical Genetics and Genomics; cHet = likely compound heterozygous; Cons. = consanguineous; Cons. carriers, n = number of variant carriers from consanguineous families; Het = heterozygous; Hom = homozygous; LP = likely pathogenic; n = number of probands with the variant; P = pathogenic; undet. = undetermined consanguinity status. EYS c.4321C>T was disease-causing (homozygous) in three unrelated probands. A fourth unrelated proband carried it heterozygously as an incidental allele (case solved by biallelic ADGRV1 c.14329C>T), and one further homozygous instance occurred in an excluded affected relative (Supplementary Table); neither is included in the causal count of three. Inheritance Pattern and Consanguinity Consanguinity exerted a statistically significant effect on zygosity within the AR-diagnosed group: all 17 consanguineous AR cases were homozygous biallelic, compared with 17 of 25 (68.0%) nonconsanguineous AR cases; the remaining 8 (32.0%) were likely compound heterozygous (P = 0.013, Fisher exact test). Overall, AR disease predominated among the 50 diagnosed probands: 44 (88.0%) harbored biallelic variants, comprising 36 homozygous and 8 likely compound-heterozygous cases (Fig 2A). Autosomal dominant disease accounted for 4 probands (8.0%), each harboring a single heterozygous P/LP variant in a known autosomal dominant gene (RHO [n = 2], NR2E3 [n = 1], and PRPF31 [n = 1]). Two male probands (4.0%) had hemizygous pathogenic variants in RPGR, consistent with X-linked disease. The single consanguineous diagnosed proband with non-AR disease carried a heterozygous Pathogenic PRPF31 variant consistent with autosomal dominant RP.Figure 2Inheritance architecture and predicted consequence classes of pathogenic and LP variants. (A) Proportional stacked horizontal bars showing inheritance pattern and zygosity among diagnosed unrelated probands, stratified by consanguinity status. The upper bar shows nonconsanguineous probands (n = 30); the lower bar shows consanguineous diagnosed probands (n = 18). Two diagnosed probands with undetermined consanguinity are excluded from this display. Categories are AR homozygous, AR likely compound heterozygous, autosomal dominant, and X-linked. Among consanguineous diagnosed probands, 17/18 (94.4%) had AR disease and all such cases were homozygous. Among nonconsanguineous probands, AR disease remained predominant (25/30, 83.3%) and comprised 17 homozygous and 8 likely compound heterozygous cases; the remaining cases were autosomal dominant (n = 3) or X-linked (n = 2). (B) Horizontal bar chart showing the consequence-class distribution of 59 distinct pathogenic and LP variants. Predicted LoF variants accounted for 45/59 (76.3%) and included stop-gained (n = 19), frameshift deletion (n = 12), frameshift duplication (n = 5), splice-site (n = 6), initiation-codon loss (n = 2), and exon-level deletion (n = 1) variants. Missense variants accounted for 13/59 (22.0%), and 1 synonymous variant was classified as LP. AR = autosomal recessive; LoF = loss-of-function; LP = likely pathogenic; P = pathogenic. Inheritance architecture and predicted consequence classes of pathogenic and LP variants. (A) Proportional stacked horizontal bars showing inheritance pattern and zygosity among diagnosed unrelated probands, stratified by consanguinity status. The upper bar shows nonconsanguineous probands (n = 30); the lower bar shows consanguineous diagnosed probands (n = 18). Two diagnosed probands with undetermined consanguinity are excluded from this display. Categories are AR homozygous, AR likely compound heterozygous, autosomal dominant, and X-linked. Among consanguineous diagnosed probands, 17/18 (94.4%) had AR disease and all such cases were homozygous. Among nonconsanguineous probands, AR disease remained predominant (25/30, 83.3%) and comprised 17 homozygous and 8 likely compound heterozygous cases; the remaining cases were autosomal dominant (n = 3) or X-linked (n = 2). (B) Horizontal bar chart showing the consequence-class distribution of 59 distinct pathogenic and LP variants. Predicted LoF variants accounted for 45/59 (76.3%) and included stop-gained (n = 19), frameshift deletion (n = 12), frameshift duplication (n = 5), splice-site (n = 6), initiation-codon loss (n = 2), and exon-level deletion (n = 1) variants. Missense variants accounted for 13/59 (22.0%), and 1 synonymous variant was classified as LP. AR = autosomal recessive; LoF = loss-of-function; LP = likely pathogenic; P = pathogenic. Pathogenic Variant Types Among the 59 distinct P/LP variants identified, predicted loss-of-function variants were predominant, accounting for 45 of 59 (76.3%) (Fig 2B). These comprised 19 stop-gained variants (32.2%), 17 frameshift variants (28.8%; 12 deletions and 5 duplications), 6 splice-site variants (10.2%; 2 canonical acceptor, 3 canonical donor, and 1 small deletion disrupting a donor site), 2 initiation-codon losses (3.4%), and 1 exon-level deletion in IMPG2 (1.7%). Missense variants comprised 13 of 59 P/LP variants (22.0%). One synonymous variant in GRN (c.1179G>A) was classified as LP (1.7%) based on predicted splicing disruption and cosegregation data. Discussion Molecular diagnosis of RP in high-burden, underrepresented populations is no longer an academic exercise: as gene-directed therapies for USH2A, RPGR, and RHO advance through clinical trials and the field moves toward variant-specific eligibility criteria, the absence of region-specific genetic data directly excludes patients from care. This inaugural ARIDE report begins to close that gap for South India. A molecular diagnosis was established in 45.9% of unrelated probands which is consistent with the 47%–76% range reported across large IRD cohorts in the United States, the United Kingdom, and Asia.22, 23, 24, 25, 26, 27 It revealed a recessive, homozygosity-enriched architecture shaped by consanguinity, 3 recurrent alleles including an established South Asian founder variant (MFSD8 c.1361T>C), and a VUS burden rooted in the underrepresentation of South Asian alleles in global genomic reference databases.15,16 The lower end of this yield range in our cohort is mechanistically explained: systematic parental sequencing for biallelic phasing was unavailable for most families, and population-specific benign variants in South Indian allele space currently lack the Genome Aggregation Database representation needed to distinguish them from rare pathogenic alleles, both tractable problems as ARIDE scales. The gene spectrum observed here diverges substantially from the most directly comparable published South Indian RP series, a panel-based study in which CERKL was the leading causal gene (21%), followed by USH2A, PROM1, RDH12, EYS, CRB1, PRPF31, and PDE6B.28 In the present cohort, CERKL is entirely absent, and RDH12, CRB1, and PRPF31 are each represented by 1 or 2 cases. Conversely, ADGRV1 (8.0%) and MFSD8 (6.0%) emerge as prominent causal genes not reported in the panel-based series. These differences reflect ≥3 distinct factors. Methodologically, targeted panels are blind to genes outside their design; ADGRV1 and MFSD8 likely escaped detection for this reason. Biologically, interpopulation heterogeneity in founder allele frequencies is well-documented across South Indian communities that have remained endogamously structured for generations.29,30 A third, underappreciated factor is enrollment design: in health systems where genetic testing is self-funded and third-party payer coverage is absent or limited, studies that recruit patients who elect and pay for testing preferentially capture higher-income, urban, outbred families, compressing the apparent contribution of consanguinity-associated genes and elevating the representation of dominant and compound heterozygous variants. The ARIDE registry enrolled participants consecutively and independently of payment capacity, making it structurally resistant to this selection bias and more likely to reflect the actual South Indian RP genetic landscape than studies whose ascertainment was payment-dependent. Taken together, these gene frequency differences argue strongly against extrapolating any single-center IRD gene catalog across the region and reinforce the need for multicenter ARIDE expansion to build a representative variant map. Although ascertained on a retinal diagnosis, this cohort included syndromic as well as nonsyndromic RP, and 2 recurrent findings illustrate why genotype must guide phenotyping beyond the eye. ADGRV1, encoding the Usher type 2C protein, was identified in 4 nonconsanguineous probands, all presenting with isolated RP. Systematic audiometric evaluation is warranted: subclinical sensorineural hearing loss may be present or emerge with time, and reclassification from nonsyndromic RP to Usher syndrome type 2C would alter counseling, surveillance, and ultimately trial eligibility as USH2A/ADGRV1-directed therapies advance. Equally notable was MFSD8 c.1361T>C (p.Met454Thr) in 3 unrelated probands, all homozygous, 2 from consanguineous families, all presenting with isolated RP without neurological features. MFSD8 encodes a lysosomal membrane protein whose biallelic truncating variants cause neuronal ceroid lipofuscinosis type 7, a severe progressive neurodegenerative disease.31,32 This allele is an established founder mutation of South Asian origin, shown to cause nonsyndromic retinal dystrophy without neurological involvement.33 Our 3 homozygous probands confirm its recurrence in a South Indian RP cohort. Given the MFSD8/ceroid lipofuscinosis type 7 spectrum, baseline neurological assessment and longitudinal follow-up remain prudent, though a retinal-limited course is the expected presentation for this genotype.34,35 Both findings illustrate why genotype–phenotype correlation is integral to this program. The broader lesson from this pilot is that deep phenotyping is essential: although our MFSD8 probands showed no neurological involvement, the same genotype can cause severe systemic disease, and other patients carrying genotypes with syndromic potential may indeed manifest extraocular features that ocular assessment alone would miss. Genotype-directed extraocular assessment, audiometry where Usher genes are implicated, neurological and biomarker evaluation where neuronal ceroid lipofuscinosis genes are implicated, and systematic deep phenotyping more generally, is therefore a defined next step for the program, both to refine diagnoses and to detect systemic disease in patients currently classified as nonsyndromic.36 Beyond these individual gene findings, consanguinity shaped the cohort's architecture at a population level: all 17 consanguineous AR cases were homozygous biallelic, compared with 17 of 25 (68.0%) in the nonconsanguineous group (P = 0.013, Fisher exact test). The diagnostic yield difference (51.4% vs. 41.7%) was directionally consistent but did not reach statistical significance. Notably, the 68.0% homozygosity rate in the nonconsanguineous AR group is higher than typically reported in many outbred cohorts, and cannot be explained by recent close-kin unions alone.37 We propose long-standing endogamous population structure, marriage within the same community or caste group over generations, independent of declared parental consanguinity, as a plausible contributing mechanism that elevates community-specific allele frequencies to the point where homozygous offspring can arise without a recent shared ancestor.29,30,38 The predominance of homozygous over heterozygous genotypes might be expected to bear on variant classification directly, but the difference in pathogenic versus LP composition between the 2 groups instead reflects the evidence base of the alleles involved. Likely pathogenic variants formed a larger share of the distinct variants among consanguineous probands (28.6% vs. 12.6% in nonconsanguineous probands), whose diagnoses more often rested on rarer, population-specific homozygous alleles whose evidence, including phenotype-concordant homozygosity and segregation, reaches LP but falls short of the independent functional, database, and literature evidence required for pathogenic. Pathogenic variants formed a larger share among nonconsanguineous probands (18.1% vs. 6.3%), reflecting recurrent alleles such as EYS c.4321C>T and ADGRV1 c.14329C>T that already carry mature ClinVar and published evidence sufficient for Pathogenic classification. Homozygosity in a phenotype-concordant proband strengthens case-level interpretation but does not by itself elevate a variant's ACMG class, which remains variant- and evidence-specific. The recurrence of EYS c.4321C>T across unrelated probands may reflect a founder effect, local allele enrichment, or ascertainment rather than a specific mutational mechanism; haplotype analysis with single-nucleotide polymorphism-array or whole-genome data and matched population controls, beyond the resolution of exome sequencing, would be required to distinguish these, and a confirmed shared haplotype could support cost-effective first-tier genotyping of this allele in South Indian RP populations. Beyond these individual recurrent alleles, the cohort carried a broad background of recessive IRD alleles. Six of the 50 diagnosed probands (12.0%) carried pathogenic or LP alleles in 2 distinct recessive IRD genes, yet only one was biallelic at both: a consanguineous proband homozygous for both a pathogenic USH2A and a LP CNGB1 variant, a dual biallelic finding of potential diagnostic relevance involving 2 recessive retinal-disease genes. In the other 5, the second allele was heterozygous, an incidental carrier finding rather than a second disease, and in one it was EYS c.4321C>T, the variant causal in the aforementioned homozygous form. A 2-gene carrier rate of this magnitude implies a high population frequency of recessive IRD alleles, an estimate that depends only on allele counts and not on any model of homozygosity. A second and weaker signal was the recurrence of homozygosity beyond the causal locus: several consanguineous probands carried additional homozygous rare variants, mostly of uncertain significance, at unlinked loci, compatible with autozygous tracts spanning >1 disease gene. Resolving such tracts genome-wide exceeds the marker density of exome data; runs-of-homozygosity analysis across the full cohort, planned, will be required to measure the extent of autozygosity and to apportion it between recent consanguinity and sustained endogamy. The corollary is interpretive: where multiple homozygous candidates coexist, a single homozygous pathogenic allele is necessary but not sufficient, and causal assignment should follow phenotype concordance rather than zygosity alone. Beyond the challenge of multiple homozygous candidates lies the larger barrier of classification itself. The 68.3% VUS fraction, the principal constraint on clinical translation, is a direct consequence of the genomic equity deficit: South Asian alleles remain substantially underrepresented in the Genome Aggregation Database and ClinVar, leaving population-specific variants at uncertain classification15,16 Resolution follows 4 ranked pathways: parental segregation testing to phase biallelic variants and establish cosegregation evidence; systematic curated deposition to ClinVar and the Clinical Genome Resource expert curation; scheduled reanalysis at a pragmatic interval of 18 months as reference databases mature; and whole-genome sequencing in persistently unresolved families to access deep intronic and structural variants beyond exome capture. The strengths of this study lie in its prospective design, consecutive enrollment independent of payment capacity, standardized ACMG/AMP adjudication with prespecified case-status rules, and multimodal imaging captured at enrollment for future phenotype–genotype analyses.21 Limitations include the absence of systematic parental sequencing for phasing and reliance on laboratory-issued WES reports without uniform orthogonal confirmation. As an exome-based assay, this study is also inherently limited in detecting variants outside captured coding regions, including deep intronic and noncoding variants, large structural and copy-number variants, repeat expansions, and mitochondrial variants; these may account for a proportion of the inconclusive and unsolved cases. Finally, single-center enrollment may not fully represent the genetic diversity of the broader South Indian RP population. Future directions follow 2 tracks, variant resolution and population-genetic analysis. We plan to resolve the VUS burden in a tiered manner: as South Asian reference data accrue, VUS will first undergo computational reanalysis and population-frequency recalibration; unresolved variants will then be assessed by parental and family-based segregation testing, for which DNA from consenting relatives has already been collected, with curated deposition to ClinVar and Clinical Genome Resource expert review and scheduled reanalysis at a pragmatic 18-month interval; variants beyond exome resolution will be pursued through genome-wide single-nucleotide polymorphism-array genotyping for copy-number and structural variants, with orthogonal confirmation, and whole-genome sequencing in persistently unresolved families for deep intronic and structural variants; and variants that remain unresolved yet clinically prioritized will undergo targeted functional validation. The same single-nucleotide polymorphism-array genotyping will additionally enable formal runs of homozygosity and identity-by-descent analysis, which, unlike exome-derived runs of homozygosity, can resolve the genome-wide segment-length distribution needed to distinguish recent consanguinity from long-standing endogamy. Variant prioritization will be further strengthened by image-based deep phenotyping, which derives gene-level predictions directly from multimodal retinal imaging and identifies phenotypically similar cases even for genes outside its training set. The standardized imaging captured across the cohort provides the substrate to validate and extend such tools in a South Asian population, and, so validated, image-based gene prediction could revolutionize access to molecular diagnosis where sequencing is self-funded, triaging and thereby extending diagnosis to the high-burden, underrepresented populations that stand to gain most. Planned multi-institutional expansion of ARIDE toward a substantially larger cohort will generate the South Indian allele-frequency data needed to resolve the VUS burden, enable autozygosity-based novel gene discovery, and, most urgently, establish a systematically genotyped, trial-ready patient pool as gene-directed therapies advance toward clinical availability in the populations that need them most.