Implementation of Artificial Intelligence-Based Diabetic Retinopathy Screening in a Tertiary Care Hospital in Quebec: Prospective Validation Study.

BACKGROUND: Diabetic retinopathy (DR) affects about 25% of people with diabetes in Canada. Early detection of DR is essential for preventing vision loss. OBJECTIVE: We evaluated the real-world performance of an artificial intelligence (AI) system that analyzes fundus images for DR screening in a Quebec tertiary care center. METHODS: We prospectively recruited adult patients with diabetes at the Centre hospitalier de l'Université de Montréal (CHUM) in Montreal, Quebec, Canada. Patients underwent dual-pathway screening: first by the Computer Assisted Retinal Analysis (CARA) AI system (index test), then by standard ophthalmological examination (reference standard). We measured the AI system's sensitivity and specificity for detecting referable disease at the patient level, along with its performance for detecting any retinopathy and diabetic macular edema (DME) at the eye level, and potential cost savings. RESULTS: This study included 115 patients. CARA demonstrated a sensitivity of 87.5% (95% CI 71.9-95.0) and specificity of 66.2% (95% CI 54.3-76.3) for detecting referable disease at the patient level. For any retinopathy detection at the eye level, CARA showed 88.2% sensitivity (95% CI 76.6-94.5) and 71.4% specificity (95% CI 63.7-78.1). For DME detection, CARA had 100% sensitivity (95% CI 64.6-100) and 81.9% specificity (95% CI 75.6-86.8). Potential yearly savings from implementing CARA at the CHUM were estimated at CAD $245,635 (US $177,643.23, as of July 26, 2024) considering 5000 patients with diabetes. CONCLUSIONS: Our study indicates that integrating a semiautomated AI system for DR screening demonstrates high sensitivity for detecting referable disease in a real-world setting. This system has the potential to improve screening efficiency and reduce costs at the CHUM, but more work is needed to validate it.

Rare Genetic Variants of Large Effect Influence Risk of Type 1 Diabetes.

Most replicated genetic determinants for type 1 diabetes are common (minor allele frequency [MAF] >5%). We aimed to identify novel rare or low-frequency (MAF <5%) single nucleotide polymorphisms with large effects on risk of type 1 diabetes. We undertook deep imputation of genotyped data followed by genome-wide association testing and meta-analysis of 9,358 type 1 diabetes case and 15,705 control subjects from 12 European cohorts. Candidate variants were replicated in a separate cohort of 4,329 case and 9,543 control subjects. Our meta-analysis identified 27 independent variants outside the MHC, among which 3 were novel and had MAF <5%. Three of these variants replicated with P replication < 0.05 and P combined < P discovery In silico analysis prioritized a rare variant at 2q24.3 (rs60587303 [C], MAF 0.5%) within the first intron of STK39, with an effect size comparable with those of common variants in the INS and PTPN22 loci (combined [from the discovery and replication cohorts] estimate of odds ratio [ORcombined] 1.97, 95% CI 1.58-2.47, P combined = 2.9 × 10-9). Pharmacological inhibition of Stk39 activity in primary murine T cells augmented effector responses through enhancement of interleukin 2 signaling. These findings provide insight into the genetic architecture of type 1 diabetes and have identified rare variants having a large effect on disease risk.

Analyses of associations with asthma in four asthma population samples from Canada and Australia.

Asthma, atopy, and related phenotypes are heterogeneous complex traits, with both genetic and environmental risk factors. Extensive research has been conducted and over hundred genes have been associated with asthma and atopy phenotypes, but many of these findings have failed to replicate in subsequent studies. To separate true associations from false positives, candidate genes need to be examined in large well-characterized samples, using standardized designs (genotyping, phenotyping and analysis). In an attempt to replicate previous associations we amalgamated the power and resources of four studies and genotyped 5,565 individuals to conduct a genetic association study of 93 previously associated candidate genes for asthma and related phenotypes using the same set of 861 single-nucleotide polymorphisms (SNPs), a common genotyping platform, and relatively harmonized phenotypes. We tested for association between SNPs and the dichotomous outcomes of asthma, atopy, atopic asthma, and airway hyperresponsiveness using a general allelic likelihood ratio test. No SNP in any gene reached significance levels that survived correction for all tested SNPs, phenotypes, and genes. Even after relaxing the usual stringent multiple testing corrections by performing a gene-based analysis (one gene at a time as if no other genes were typed) and by stratifying SNPs based on their prior evidence of association, no genes gave strong evidence of replication. There was weak evidence to implicate the following: IL13, IFNGR2, EDN1, and VDR in asthma; IL18, TBXA2R, IFNGR2, and VDR in atopy; TLR9, TBXA2R, VDR, NOD2, and STAT6 in airway hyperresponsiveness; TLR10, IFNGR2, STAT6, VDR, and NPSR1 in atopic asthma. Additionally we found an excess of SNPs with small effect sizes (OR < 1.4). The low rate of replication may be due to small effect size, differences in phenotypic definition, differential environmental effects, and/or genetic heterogeneity. To aid in future replication studies of asthma genes a comprehensive database was compiled and is available to the scientific community at http://genapha.icapture.ubc.ca/.

Effects of DNA supercoiling and topoisomerases on the expression of genes coding for F165(1), a P-like fimbriae.

Pathogenic Escherichia coli 4787 (O115:KV165) causes septicemia in pigs and expresses the fimbriae F165(1) encoded by the foo operon that belongs to the P fimbrial family. fooI and fooB, encoding specific foo regulators, are divergently transcribed; their intergenic region is responsible for the regulation of foo expression. The role of global and local supercoiling (transcription-induced supercoiling within the intergenic region) on the regulation of foo expression was investigated. Expression of fooB was significantly altered when global negative supercoiling was reduced by a mutation that decreases DNA gyrase activity. Deletion of the topA gene, encoding for topoisomerase I that relaxes local negative supercoiling, further reduced fooB expression. This suggests that both global and local supercoiling can significantly affect fooB expression. Moreover, FooI, a positive regulator of fooB expression, has no effect on fooB expression in the topA null mutant. This study showed that divergent transcription from a strong promoter can significantly enhance fooB expression and compensate for the absence of FooI in a wild-type strain.

Genetic control of alternative splicing in the TAP2 gene: possible implication in the genetics of type 1 diabetes.

The transporter 2, ATP-binding cassette, subfamily B (TAP2) is involved in the transport of antigenic peptides to HLA molecules. Coding TAP2 polymorphisms shows a strong association with type 1 diabetes, but it is not clear whether this association may be entirely due to linkage disequilibrium with HLA DR and DQ. Functionally, rat Tap2 nonsynonymous single-nucleotide polymorphisms (nsSNPs) confer differential selectivity for antigenic peptides, but this was not shown to be the case for human TAP2 nsSNPs. In the human, differential peptide selectivity is rather conferred by two splicing isoforms with alternative carboxy terminals. Here, we tested the hypothesis that alleles at the coding SNPs favor different splicing isoforms, thus determining peptide selectivity indirectly. This may be the basis for independent contribution to the type 1 diabetes association. In RNA from heterozygous lymphoblastoid lines, we measured the relative abundance of each SNP haplotype in each isoform. In isoform NM_000544, the G (Ala) allele at 665 Thr>Ala (rs241447) is more than twice as abundant as A (Thr) (GA = 2.2 +/- 0.4, P = 1.5 x 10(-4)), while isoform NM_018833 is derived almost exclusively from chromosomes carrying A (AG = 18.1 +/- 5.6, P = 2.04 x 10(-7)). In 889 Canadian children with type 1 diabetes, differential transmission of parental TAP2 alleles persisted (P = 0.011) when analysis was confined to chromosomes carrying only DQ*02 alleles, which mark a conserved DR-DQ haplotype, thus eliminating most of the variation at DR-DQ. Thus, we present evidence of TAP2 association with type 1 diabetes that is independent of HLA DR-DQ and describe a plausible functional mechanism based on allele dependence of splicing into isoforms known to have differential peptide selectivities.

Association of the cytotoxic T lymphocyte-associated antigen 4 gene with type 1 diabetes: evidence for independent effects of two polymorphisms on the same haplotype block.

A recent study mapped the known association of type 1 diabetes with the cytotoxic T lymphocyte-associated antigen 4 gene to a polymorphism at the 3'end (+6230G>A), but could not rule out additional contribution from the 5' end of the gene. To examine this possibility, we analyzed four polymorphisms at the 5'-flanking region for effects independent of +6230G>A. We confirm, by the transmission disequilibrium test, in 496 family trios overtransmission of the susceptibility allele (G) at +6230 (217/168; P = 0.013). Of the four promoter polymorphisms, one (-319C>T) showed overtransmission of the C allele (97/58; P = 0.0017). Because the undertransmitted T at the promoter is in linkage disequilibrium with the overtransmitted G at +6230G>A, the effect observed at the promoter cannot be accounted for by linkage disequilibrium with the +6230G>A. We confirm this by showing that parents heterozygous at the promoter but homozygous at +6230 overtransmit the C promoter allele even more significantly (53/24; P = 9 x 10(-4)). In vitro, the T promoter allele directs higher luciferase expression in Jurkat cells by 42% (P = 0.006), a difference also found in lymphocyte mRNA from eight individuals heterozygous at the promoter, but homozygous at +6230 (P = 1.3 x 10(-4)). Thus, the +6230G>A cannot be the sole functional variant. Either the two polymorphisms define a haplotype carrying the (yet unexamined) functional variant or the -319C>T contributes to the genetic association independently, a possibility suggested by the functional evidence we present.

Influence of environmental cues on transcriptional regulation of foo and clp coding for F165(1) and CS31A adhesins in Escherichia coli.

F165(1) (foo) and CS31A (clp) are bacterial adhesins synthesized by Escherichia coli strains associated with diarrhea and septicemia in piglets and calves. They belong to the P-regulatory family and as such are subject to a phase variation control mediated by Lrp (leucine responsive regulatory protein) and regulators homologous to PapI. Analysis of expression of transcriptional fusions between the fooB or fooI promoters and lacZ showed that Lrp is an activator of foo and fooI transcription, whereas it represses clp transcription. Furthermore, foo phase variation leads to a large majority of phase-ON cells, whereas clp phase variation leads to a majority of phase-OFF cells. We compared the influence of several environmental cues on foo and clp expression, with special attention to the effects of leucine and alanine known to be mediated by Lrp. Inhibition or significant repression of foo and clp transcription was observed at low temperature, in LB medium, and in the presence of glucose, alanine, or leucine. Glucose repression of foo but not of clp was totally relieved by addition of cAMP. Osmolarity and pH had little effect. Alanine but not leucine, and LB medium inhibited foo and clp phase variation, locking cells in the OFF phase. Low temperature inhibited clp phase variation and altered the switch frequency of foo phase variation, leading to more phase-OFF cells. Glucose altered the phase variation of both operons, increasing the number of phase-OFF cells in the population. The regulation pattern of foo and clp is consistent with F165(1) and CS31A production in low nutrient environments, even at moderately acidic pH or high osmolarity.

A common autoimmunity predisposing signal peptide variant of the cytotoxic T-lymphocyte antigen 4 results in inefficient glycosylation of the susceptibility allele.

A common T17A polymorphism in the signal peptide of the cytotoxic T-lymphocyte antigen 4 (CTLA-4), a T-cell receptor that negatively regulates immune responses, is associated with risk for autoimmune disease. Because the polymorphism is absent from the mature protein, we hypothesized that its biological effect must involve early stages of protein processing, prior to signal peptide cleavage. Constructs representing the two alleles were compared by in vitro translation, in the presence of endoplasmic reticulum membranes. We studied glycosylation by endoglycosidase H digestion and glycosylation mutant constructs, cleavage of peptide with inhibitors, and membrane integration by ultracentrifugation and proteinase K sensitivity. A major cleaved and glycosylated product was seen for both alleles of the protein but a band representing incomplete glycosylation was markedly more abundant in the predisposing Ala allele (32.7 +/- 1.0 versus 10.6% +/- 1.2 for Thr, p < 10(-9)). In addition, differential intracellular/surface partitioning was studied with co-transfection of the alleles fused to distinct fluorescent proteins in COS-1 cells. By quantitative confocal microscopy we found a higher ratio of cell surface/total CTLAThr(17) versus CTLAAla(17) (p = 0.01). Our findings corroborate observations, in other proteins, that the signal peptide can determine the efficiency of post-translational modifications other than cleavage and suggest inefficient processing of the autoimmunity predisposing Ala allele as the explanation for the genetic effect.