Age-related macular degeneration (AMD) is a common threat to vision. While classification of disease stages is critical to understanding disease risk and progression, several systems based on color ...fundus photographs are known. Most of these require in-depth and time-consuming analysis of fundus images. Herein, we present an automated computer-based classification algorithm.
Algorithm development for AMD classification based on a large collection of color fundus images. Validation is performed on a cross-sectional, population-based study.
We included 120 656 manually graded color fundus images from 3654 Age-Related Eye Disease Study (AREDS) participants. AREDS participants were >55 years of age, and non-AMD sight-threatening diseases were excluded at recruitment. In addition, performance of our algorithm was evaluated in 5555 fundus images from the population-based Kooperative Gesundheitsforschung in der Region Augsburg (KORA; Cooperative Health Research in the Region of Augsburg) study.
We defined 13 classes (9 AREDS steps, 3 late AMD stages, and 1 for ungradable images) and trained several convolution deep learning architectures. An ensemble of network architectures improved prediction accuracy. An independent dataset was used to evaluate the performance of our algorithm in a population-based study.
κ Statistics and accuracy to evaluate the concordance between predicted and expert human grader classification.
A network ensemble of 6 different neural net architectures predicted the 13 classes in the AREDS test set with a quadratic weighted κ of 92% (95% confidence interval, 89%–92%) and an overall accuracy of 63.3%. In the independent KORA dataset, images wrongly classified as AMD were mainly the result of a macular reflex observed in young individuals. By restricting the KORA analysis to individuals >55 years of age and prior exclusion of other retinopathies, the weighted and unweighted κ increased to 50% and 63%, respectively. Importantly, the algorithm detected 84.2% of all fundus images with definite signs of early or late AMD. Overall, 94.3% of healthy fundus images were classified correctly.
Our deep learning algoritm revealed a weighted κ outperforming human graders in the AREDS study and is suitable to classify AMD fundus images in other datasets using individuals >55 years of age.
Mitochondrial (MT) dysfunction is a hallmark of aging and has been associated with most aging-related diseases as well as immunological processes. However, little is known about aging, lifestyle and ...genetic factors influencing mitochondrial DNA (mtDNA) abundance. In this study, mtDNA abundance was estimated from the weighted intensities of probes mapping to the MT genome in 295,150 participants from the UK Biobank. We found that the abundance of mtDNA was significantly elevated in women compared to men, was negatively correlated with advanced age, higher smoking exposure, greater body-mass index, higher frailty index as well as elevated red and white blood cell count and lower mortality. In addition, several biochemistry markers in blood-related to cholesterol metabolism, ion homeostasis and kidney function were found to be significantly associated with mtDNA abundance. By performing a genome-wide association study, we identified 50 independent regions genome-wide significantly associated with mtDNA abundance which harbour multiple genes involved in the immune system, cancer as well as mitochondrial function. Using mixed effects models, we estimated the SNP-heritability of mtDNA abundance to be around 8%. To investigate the consequence of altered mtDNA abundance, we performed a phenome-wide association study and found that mtDNA abundance is involved in risk for leukaemia, hematologic diseases as well as hypertension. Thus, estimating mtDNA abundance from genotyping arrays has the potential to provide novel insights into age- and disease-relevant processes, particularly those related to immunity and established mitochondrial functions.
Genes involved in cancer are under constant evolutionary pressure, potentially resulting in diverse molecular properties. In this study, we explore 23 omic features from publicly available databases ...to define the molecular profile of different classes of cancer genes. Cancer genes were grouped according to mutational landscape (germline and somatically mutated genes), role in cancer initiation (cancer driver genes) or cancer survival (survival genes), as well as being implicated by genome-wide association studies (GWAS genes). For each gene, we also computed feature scores based on all omic features, effectively summarizing how closely a gene resembles cancer genes of the respective class. In general, cancer genes are longer, have a lower GC content, have more isoforms with shorter exons, are expressed in more tissues and have more transcription factor binding sites than non-cancer genes. We found that germline genes more closely resemble single tissue GWAS genes while somatic genes are more similar to pleiotropic cancer GWAS genes. As a proof-of-principle, we utilized aggregated feature scores to prioritize genes in breast cancer GWAS loci and found that top ranking genes were enriched in cancer related pathways. In conclusion, we have identified multiple omic features associated with different classes of cancer genes, which can assist prioritization of genes in cancer gene discovery.
Patients with active nonvariceal upper gastrointestinal bleeding (NVUGIB) usually require urgent endoscopic treatment. Standard therapy (ST) using haemoclip + / - epinephrine injection is not always ...successful. Bipolar haemostatic forceps (HemoStat/Pentax®) are an approved medical device for the treatment of gastrointestinal bleeding. However, their use as a primary endoscopic treatment for active NVUGIB has not yet been proven in a randomized prospective study.
This is a prospective, randomized, multicentre superiority trial (n ≥ 5). Patients with active NVUGIB will be randomized (1:1) to ST and to experimental therapy (ET) by application of bipolar haemostatic forceps. In the case of failed initial treatment within 15 min, crossover treatment will be attempted first. Rescue treatment (e.g. via over-the-scope-clip) will then be allowed after 30 min. All patients will also receive standard therapy with proton pump inhibitors. Forty-five patients per treatment arm are required to demonstrate an absolute difference of 25.4% with a power of 80% and a significance level of 0.05.
The hypothesis of the study is that bipolar haemostatic forceps are superior to ST in terms of successful primary haemostasis and the absence of recurrent bleeding within 30 days (combined endpoint). The 1:1 randomization is also ethically justifiable for this study, as both procedures are approved for the intervention in question. To further increase the safety of the patients in the study, crossover treatment and rescue treatment are planned. The prospective design seems feasible in a reasonable time frame (recruitment period of 12 months), as nonvariceal upper gastrointestinal bleeding is common. Anticoagulants and/or antiplatelet drugs could be an important confounding factor in the statistical analysis that needs to be taken into account and calculated if necessary. In conclusion, this randomized, prospective, multicentre study could make an important contribution to answering the question of whether bipolar haemostatic forceps could be the first-line therapy in the endoscopic treatment of stage Forrest I a + b NVUGIB.
ClinicalTrials.gov NCT05353062. Registered on April 30 2022.
Breast cancer (BC) patients diagnosed between two screenings (interval cancers) are more likely than screen-detected patients to carry rare deleterious mutations in cancer genes potentially leading ...to increased risk for other non-breast cancer (non-BC) tumors. In this study, we include 14,846 women diagnosed with BC of which 1,772 are interval and 13,074 screen-detected. Compared to women with screen-detected cancers, interval breast cancer patients are more likely to have a non-BC tumor before (Odds ratio (OR): 1.43 1.19-1.70, P = 9.4 x 10
) and after (OR: 1.28 1.14-1.44, P = 4.70 x 10
) breast cancer diagnosis, are more likely to report a family history of non-BC tumors and have a lower genetic risk score based on common variants for non-BC tumors. In conclusion, interval breast cancer is associated with other tumors and common cancer variants are unlikely to be responsible for this association. These findings could have implications for future screening and prevention programs.
Stargardt disease (STGD1) is an autosomal recessive retinopathy, caused by mutations in the retina-specific ATP-binding cassette transporter (ABCA4) gene. To establish the mutational spectrum and to ...assess effects of selected deep intronic and common genetic variants on disease, we performed a comprehensive sequence analysis in a large cohort of German STGD1 patients.
DNA samples of 335 STGD1 patients were analyzed for ABCA4 mutations in its 50 coding exons and adjacent intronic sequences by resequencing array technology or next generation sequencing (NGS). Parts of intron 30 and 36 were screened by Sanger chain-terminating dideoxynucleotide sequencing. An in vitro splicing assay was used to test selected variants for their splicing behavior. By logistic regression analysis we assessed the association of common ABCA4 alleles while a multivariate logistic regression model calculated a genetic risk score (GRS).
Our analysis identified 148 pathogenic or likely pathogenic mutations, of which 48 constitute so far unpublished ABCA4-associated disease alleles. Four rare deep intronic variants were found once in 472 alleles analyzed. In addition, we identified six risk-modulating common variants. Genetic risk score estimates suggest that defined common ABCA4 variants influence disease risk in carriers of a single pathogenic ABCA4 allele.
Our study adds to the mutational spectrum of the ABCA4 gene. Moreover, in our cohort, deep intronic variants in intron 30 and 36 likely play no or only a minor role in disease pathology. Of note, our findings demonstrate a possible modifying effect of common sequence variants on ABCA4-associated disease.
There is a rising concern about treatment-associated cardiotoxicities in breast cancer patients. This study aimed to determine the time- and treatment-specific incidence of arrhythmia, heart failure, ...and ischemic heart disease in women diagnosed with breast cancer.
A register-based matched cohort study was conducted including 8015 breast cancer patients diagnosed from 2001 to 2008 in the Stockholm-Gotland region and followed up until 2017. Time-dependent risks of arrhythmia, heart failure, and ischemic heart disease in breast cancer patients were assessed using flexible parametric models as compared to matched controls from general population. Treatment-specific effects were estimated in breast cancer patients using Cox model.
Time-dependent analyses revealed long-term increased risks of arrhythmia and heart failure following breast cancer diagnosis. Hazard ratios (HRs) within the first year of diagnosis were 2.14 (95% CI = 1.63-2.81) for arrhythmia and 2.71 (95% CI = 1.70-4.33) for heart failure. HR more than 10 years following diagnosis was 1.42 (95% CI = 1.21-1.67) for arrhythmia and 1.28 (95% CI = 1.03-1.59) for heart failure. The risk for ischemic heart disease was significantly increased only during the first year after diagnosis (HR = 1.45, 95% CI = 1.03-2.04). Trastuzumab and anthracyclines were associated with increased risk of heart failure. Aromatase inhibitors, but not tamoxifen, were associated with risk of ischemic heart disease. No increased risk of heart disease was identified following locoregional radiotherapy.
Administration of systemic adjuvant therapies appears to be associated with increased risks of heart disease. The risk estimates observed in this study may aid adjuvant therapy decision-making and patient counseling in oncology practices.
This work was supported by the Swedish Research Council (grant no: 2018-02547); Swedish Cancer Society (grant no: CAN-19-0266); and FORTE (grant no: 2016-00081).
Using exome sequencing, the underlying variants in many persons with autosomal recessive diseases remain undetected. We explored autosomal recessive Stargardt disease (STGD1) as a model to identify ...the missing heritability.
Sequencing of ABCA4 was performed in 8 STGD1 cases with one variant and p.Asn1868Ile in trans, 25 cases with one variant, and 3 cases with no ABCA4 variant. The effect of intronic variants was analyzed using in vitro splice assays in HEK293T cells and patient-derived fibroblasts. Antisense oligonucleotides were used to correct splice defects.
In 24 of the probands (67%), one known and five novel deep-intronic variants were found. The five novel variants resulted in messenger RNA pseudoexon inclusions, due to strengthening of cryptic splice sites or by disrupting a splicing silencer motif. Variant c.769-784C>T showed partial insertion of a pseudoexon and was found in cis with c.5603A>T (p.Asn1868Ile), so its causal role could not be fully established. Variant c.4253+43G>A resulted in partial skipping of exon 28. Remarkably, antisense oligonucleotides targeting the aberrant splice processes resulted in (partial) correction of all splicing defects.
Our data demonstrate the importance of assessing noncoding variants in genetic diseases, and show the great potential of splice modulation therapy for deep-intronic variants.
The extent of aneuploidy of the sex chromosomes increases with age in human leukocytes. Here, we re-explore the dynamics of normal loss of the Y chromosome (
LOY
) with age based on microarray data ...using two exponential models and two different ways to estimate the fraction of
LOY
. This analysis shows the existence of a significant correlation between the fraction of
LOY
estimated from molecular cytogenetics and genotyping microarray data. Although the specific estimates of the parameters for the two exponential models are different from those derived from cytogenetics data, the present analysis in an independent dataset of normal individuals confirms that X0 cells have a selective advantage over XY cells. Moreover, patients with age-related macular degeneration display higher fraction of
LOY
values and seem to have a predisposition to lose their Y chromosome even at young ages compared to control individuals. As there are no data available for the same individuals at different time points, the parameters reported here are average values drawn from population analyses.