Generative pretrained transformer-4, an artificial intelligence text predictive model, has a high capability for passing novel written radiology exam questions.

PURPOSE: AI-image interpretation, through convolutional neural networks, shows increasing capability within radiology. These models have achieved impressive performance in specific tasks within controlled settings, but possess inherent limitations, such as the inability to consider clinical context. We assess the ability of large language models (LLMs) within the context of radiology specialty exams to determine whether they can evaluate relevant clinical information. METHODS: A database of questions was created with official sample, author written, and textbook questions based on the Royal College of Radiology (United Kingdom) FRCR 2A and American Board of Radiology (ABR) Certifying examinations. The questions were input into the Generative Pretrained Transformer (GPT) versions 3 and 4, with prompting to answer the questions. RESULTS: One thousand seventy-two questions were evaluated by GPT-3 and GPT-4. 495 (46.2%) were for the FRCR 2A and 577 (53.8%) were for the ABR exam. There were 890 single best answers (SBA), and 182 true/false questions. GPT-4 was correct in 629/890 (70.7%) SBA and 151/182 (83.0%) true/false questions. There was no degradation on author written questions. GPT-4 performed significantly better than GPT-3 which selected the correct answer in 282/890 (31.7%) SBA and 111/182 (61.0%) true/false questions. Performance of GPT-4 was similar across both examinations for all categories of question. CONCLUSION: The newest generation of LLMs, GPT-4, demonstrates high capability in answering radiology exam questions. It shows marked improvement from GPT-3, suggesting further improvements in accuracy are possible. Further research is needed to explore the clinical applicability of these AI models in real-world settings.

Multiscale spatial mapping of cell populations across anatomical sites in healthy human skin and basal cell carcinoma.

Our understanding of how human skin cells differ according to anatomical site and tumour formation is limited. To address this, we have created a multiscale spatial atlas of healthy skin and basal cell carcinoma (BCC), incorporating in vivo optical coherence tomography, single-cell RNA sequencing, spatial global transcriptional profiling, and in situ sequencing. Computational spatial deconvolution and projection revealed the localisation of distinct cell populations to specific tissue contexts. Although cell populations were conserved between healthy anatomical sites and in BCC, mesenchymal cell populations including fibroblasts and pericytes retained signatures of developmental origin. Spatial profiling and in silico lineage tracing support a hair follicle origin for BCC and demonstrate that cancer-associated fibroblasts are an expansion of a POSTN+ subpopulation associated with hair follicles in healthy skin. RGS5+ pericytes are also expanded in BCC suggesting a role in vascular remodelling. We propose that the identity of mesenchymal cell populations is regulated by signals emanating from adjacent structures and that these signals are repurposed to promote the expansion of skin cancer stroma. The resource we have created is publicly available in an interactive format for the research community.

Systematic review of deep learning image analyses for the diagnosis and monitoring of skin disease.

Skin diseases affect one-third of the global population, posing a major healthcare burden. Deep learning may optimise healthcare workflows through processing skin images via neural networks to make predictions. A focus of deep learning research is skin lesion triage to detect cancer, but this may not translate to the wider scope of >2000 other skin diseases. We searched for studies applying deep learning to skin images, excluding benign/malignant lesions (1/1/2000-23/6/2022, PROSPERO CRD42022309935). The primary outcome was accuracy of deep learning algorithms in disease diagnosis or severity assessment. We modified QUADAS-2 for quality assessment. Of 13,857 references identified, 64 were included. The most studied diseases were acne, psoriasis, eczema, rosacea, vitiligo, urticaria. Deep learning algorithms had high specificity and variable sensitivity in diagnosing these conditions. Accuracy of algorithms in diagnosing acne (median 94%, IQR 86-98; n = 11), rosacea (94%, 90-97; n = 4), eczema (93%, 90-99; n = 9) and psoriasis (89%, 78-92; n = 8) was high. Accuracy for grading severity was highest for psoriasis (range 93-100%, n = 2), eczema (88%, n = 1), and acne (67-86%, n = 4). However, 59 (92%) studies had high risk-of-bias judgements and 62 (97%) had high-level applicability concerns. Only 12 (19%) reported participant ethnicity/skin type. Twenty-four (37.5%) evaluated the algorithm in an independent dataset, clinical setting or prospectively. These data indicate potential of deep learning image analysis in diagnosing and monitoring common skin diseases. Current research has important methodological/reporting limitations. Real-world, prospectively-acquired image datasets with external validation/testing will advance deep learning beyond the current experimental phase towards clinically-useful tools to mitigate rising health and cost impacts of skin disease.

A noninvasive method for whole-genome skin methylome profiling.

BACKGROUND: Ageing, disease and malignant transformation of the skin are associated with changes in DNA methylation. So far, mostly invasive methodologies such as biopsies have been applied in collecting DNA methylation signatures. Tape stripping offers a noninvasive option for skin diagnostics. It enables the easy but robust capture of biologic material in large numbers of participants without the need for specialized medical personnel. OBJECTIVES: To design and validate a methodology for noninvasive skin sample collection using tape stripping for subsequent DNA -methylation analysis. METHODS: A total of 175 participants were recruited and provided tape-stripping samples from a sun-exposed area; 92 provided matched tape-stripping samples from a sun-protected area, and an additional 5 provided matched skin-shave biopsies from the same area. Using -enzymatic conversion and whole-genome Illumina sequencing, we generated genome-wide DNA methylation profiles that were used to evaluate the feasibility of noninvasive data acquisition, to compare with established sampling approaches and to investigate biomarker identification for age and ultraviolet (UV) exposure. RESULTS: We found that tape-stripping samples showed strong concordance in their global DNA methylation landscapes to those of conventional invasive biopsies. Moreover, we showed sample reproducibility and consistent global methylation profiles in skin tape-stripping samples collected from different areas of the body. Using matched samples from sun-protected and sun-exposed areas of the body we were able to validate the capacity of our method to capture the effects of environmental changes and ageing in a cohort covering various ages, ethnicities and skin types. We found DNA methylation changes on the skin resulting from UV exposure and identified significant age-related hypermethylation of CpG islands, with a pronounced peak effect at 50-55 years of age, including methylation changes in well-described markers of ageing. CONCLUSIONS: These data demonstrate the feasibility of using tape stripping combined with whole-genome sequencing as a noninvasive approach to measuring DNA methylation changes in the skin. In addition, they outline a viable experimental framework for the use of skin tape stripping, particularly when it is performed in large cohorts of patients to identify biomarkers of skin ageing, UV damage and, possibly, to track treatment response to therapeutic interventions.

Fibroblast Heterogeneity in Healthy and Wounded Skin.

Fibroblasts are the main cell type in the dermis. They are responsible for the synthesis and deposition of structural proteins such as collagen and elastin, which are integrated into the extracellular matrix (ECM). Mouse and human studies using flow cytometry, cell culture, skin reconstitution, and lineage tracing experiments have shown the existence of different subpopulations of fibroblasts, including papillary fibroblasts, reticular fibroblasts, and fibroblasts comprising the dermal papilla at the base of the hair follicle. In recent years, the technological advances in single-cell sequencing have allowed researchers to study the repertoire of cells present in full-thickness skin including the dermis. Multiple groups have confirmed that distinct fibroblast populations can be identified in mouse and human dermis on the basis of differences in the transcriptional profile. Here, we discuss the current state of knowledge regarding dermal fibroblast heterogeneity in healthy mouse and human skin, highlighting the similarities and differences between mouse and human fibroblast subpopulations. We also discuss how fibroblast heterogeneity may provide insights into physiological wound healing and its dysfunction in pathological states such as hypertrophic and keloid scars.

Male genital lichen sclerosus, microincontinence and occlusion: mapping the disease across the prepuce.

BACKGROUND: Male genital lichen sclerosus (MGLSc) can lead to significant sexual dysfunction and urological morbidity, and is also a risk factor for premalignant disease (penile intraepithelial neoplasia and penile cancer), particularly squamous cell carcinoma. Although the precise aetiopathogenesis of MGLSc remains controversial, accumulated evidence indicates that it is related to chronic, intermittent, occluded exposure to urine. AIM: To perform spatial mapping of MGLSc across the human prepuce and assess how this supports the urinary occlusion hypothesis. METHODS: Preputial samples were collected from 10 patients with clinically diagnosed MGLSc undergoing circumcision. The samples were then divided into a grid pattern and 10 punch biopsies were obtained from each section to determine the extent and distribution of the disease process across each prepuce. RESULTS: All 10 patients reported having urinary microincontinence, and all were histologically confirmed as having MGLSc. The most proximal aspect of the prepuce was found to be universally affected by MGLSc in all patients, whereas the most distal part was overwhelmingly shown to be the least affected area. Of the 63 MGLSc-affected regions, 62 were in direct physical contiguity with one another. The histological extent of the disease was not found to be congruent with either the severity of the symptoms reported by the patients or the clinical examination. CONCLUSION: In uncircumcised men with urinary microincontinence, after the prepuce has been replaced post micturition, small amounts of urine can pool between the juxtaposed epithelial surfaces. The proximal aspect of the prepuce is subjected to the maximum amount of occlusion and maximal contact with accumulated urine, whereas the distal prepuce is subjected to the least. Our findings suggest that accentuated contact between urine and susceptible penile epithelium due to occlusion can lead to MGLSc. Furthermore, contiguity data suggest that once established, it is possible that MGLSc advances across tissues by physical contact. This is the first study examining the changes in the preputial landscape in patients with LSc and contributes to our understanding of disease aetiology and progression.

Angiotensin-Converting Enzyme 2 Expression Is Detectable in Keratinocytes, Cutaneous Appendages, and Blood Vessels by Multiplex RNA In Situ Hybridization.

OBJECTIVE: The angiotensin-converting enzyme 2 (ACE2) receptor mediates uptake of SARS-CoV-2, the virus responsible for COVID-19. Previous work analyzing publicly available bulk RNA-sequencing data sets has shown the expression of ACE2 in human keratinocytes. This finding is potentially relevant for the etiology of COVID-19-associated rashes and might also suggest a possible entry mechanism for the SARS-CoV-2 virus. In this study, the authors examined the spatial localization of ACE2 mRNA in vivo. METHODS AND RESULTS: The authors analyzed several publicly available single-cell RNA-sequencing data sets. They determined spatial localization of ACE2 mRNA using multiplex RNA in situ hybridization in human skin. CONCLUSIONS: Both analyses supported ACE2 expression in keratinocytes and skin vasculature, which could reflect a potential cutaneous entry point for SARS-CoV-2, particularly in damaged or broken skin. Moreover, ACE2 expression in vascular endothelial cells may support direct, virally mediated mechanisms in the etiology of the chilblain-like acral eruption that is seen in patients with COVID-19.

Gradient boosting approaches can outperform logistic regression for risk prediction in cutaneous allergy.

BACKGROUND: Contact allergy is a major clinical and public health challenge. It is important to identify individuals who are at risk and perform patch testing to identify relevant allergens. Predicting clinical risk on the basis of input parameters is common in clinical medicine and traditionally has been achieved with linear models. OBJECTIVES: We hypothesized that the risk of a clinically relevant positive patch test could be predicted according to clinical and demographic parameters. METHODS: We compared the predictive accuracy of logistic regression with more sophisticated machine learning approaches such as gradient boosting, in the prediction of patch testing results. RESULTS: We found that both logistic regression and more sophisticated machine learning approaches were able to predict the risk of positive patch tests. For certain predictions, including the overall risk of a clinically relevant positive patch test, gradient boosting approaches can outperform logistic regression. CONCLUSIONS: These findings suggest that complex nonlinear interactions between input variables are relevant in risk prediction. While a risk prediction model cannot replace the judgment of an experienced clinician, quantifying the risk of a clinically relevant positive patch test result has the potential to assist in decision making and to inform discussions with patients.

Developmental cell programs are co-opted in inflammatory skin disease.

The skin confers biophysical and immunological protection through a complex cellular network established early in embryonic development. We profiled the transcriptomes of more than 500,000 single cells from developing human fetal skin, healthy adult skin, and adult skin with atopic dermatitis and psoriasis. We leveraged these datasets to compare cell states across development, homeostasis, and disease. Our analysis revealed an enrichment of innate immune cells in skin during the first trimester and clonal expansion of disease-associated lymphocytes in atopic dermatitis and psoriasis. We uncovered and validated in situ a reemergence of prenatal vascular endothelial cell and macrophage cellular programs in atopic dermatitis and psoriasis lesional skin. These data illustrate the dynamism of cutaneous immunity and provide opportunities for targeting pathological developmental programs in inflammatory skin diseases.

Allergic contact dermatitis in atopic individuals: Results of a 30-year retrospective study.

BACKGROUND: The association between atopic dermatitis (AD) and contact allergy, remains unclear, with studies to date showing conflicting results. OBJECTIVES: To investigate the prevalence of contact allergy in AD individuals compared to those without AD. METHODS: Results of 46 250 patients patch tested in a single centre over a span of 30 years were reviewed, comparing those with AD with those without AD. Collected data were analysed with corrections for multiple confounding variables, including date of patch testing to account for changes in allergens tested over the period. RESULTS: Nine allergens showed a statistically significant difference between the two groups. Contact allergy to nickel, cobalt and primin was less likely to arise in those with AD, whilst substances found in topical dermatological products were more likely to be associated with AD. CONCLUSIONS: This is the largest single centre study of contact sensitization in atopy reported to date. The previously reported association between contact allergy to sesquiterpene lactone and AD is reinforced. The decreased incidence of metal allergy suggests distinct immunological effector mechanisms in sensitization to these allergens. In keeping with previous publications, exposure to topical treatments for AD can result in sensitisation and contact allergy and clinicians should consider patch testing in AD individuals who report worsening of their skin despite continued treatment with topical medicaments.

CpG binding protein (CFP1) occupies open chromatin regions of active genes, including enhancers and non-CpG islands.

BACKGROUND: The mechanism by which protein complexes interact to regulate the deposition of post-translational modifications of histones remains poorly understood. This is particularly important at regulatory regions, such as CpG islands (CGIs), which are known to recruit Trithorax (TrxG) and Polycomb group proteins. The CxxC zinc finger protein 1 (CFP1, also known as CGBP) is a subunit of the TrxG SET1 protein complex, a major catalyst of trimethylation of H3K4 (H3K4me3). RESULTS: Here, we used ChIP followed by high-throughput sequencing (ChIP-seq) to analyse genomic occupancy of CFP1 in two human haematopoietic cell types. We demonstrate that CFP1 occupies CGIs associated with active transcription start sites (TSSs), and is mutually exclusive with H3K27 trimethylation (H3K27me3), a marker of polycomb repressive complex 2. Strikingly, rather than being restricted to active CGI TSSs, CFP1 also occupies a substantial fraction of active non-CGI TSSs and enhancers of transcribed genes. However, relative to other TrxG subunits, CFP1 was specialised to TSSs. Finally, we found enrichment of CpG-containing DNA motifs in CFP1 peaks at CGI promoters. CONCLUSIONS: We found that CFP1 is not solely recruited to CpG islands as it was originally defined, but also other regions including non-CpG island promoters and enhancers.

Spatial and Single-Cell Transcriptional Profiling Identifies Functionally Distinct Human Dermal Fibroblast Subpopulations.

Previous studies have shown that mouse dermis is composed of functionally distinct fibroblast lineages. To explore the extent of fibroblast heterogeneity in human skin, we used a combination of comparative spatial transcriptional profiling of human and mouse dermis and single-cell transcriptional profiling of human dermal fibroblasts. We show that there are at least four distinct fibroblast populations in adult human skin, not all of which are spatially segregated. We define markers permitting their isolation and show that although marker expression is lost in culture, different fibroblast subpopulations retain distinct functionality in terms of Wnt signaling, responsiveness to IFN-γ, and ability to support human epidermal reconstitution when introduced into decellularized dermis. These findings suggest that ex vivo expansion or in vivo ablation of specific fibroblast subpopulations may have therapeutic applications in wound healing and diseases characterized by excessive fibrosis.

Fibroblast heterogeneity: implications for human disease.

Fibroblasts synthesize the extracellular matrix of connective tissue and play an essential role in maintaining the structural integrity of most tissues. Researchers have long suspected that fibroblasts exhibit functional specialization according to their organ of origin, body site, and spatial location. In recent years, a number of approaches have revealed the existence of fibroblast subtypes in mice. Here, we discuss fibroblast heterogeneity with a focus on the mammalian dermis, which has proven an accessible and tractable system for the dissection of these relationships. We begin by considering differences in fibroblast identity according to anatomical site of origin. Subsequently, we discuss new results relating to the existence of multiple fibroblast subtypes within the mouse dermis. We consider the developmental origin of fibroblasts and how this influences heterogeneity and lineage restriction. We discuss the mechanisms by which fibroblast heterogeneity arises, including intrinsic specification by transcriptional regulatory networks and epigenetic factors in combination with extrinsic effects of the spatial context within tissue. Finally, we discuss how fibroblast heterogeneity may provide insights into pathological states including wound healing, fibrotic diseases, and aging. Our evolving understanding suggests that ex vivo expansion or in vivo inhibition of specific fibroblast subtypes may have important therapeutic applications.

DNA methylation of intragenic CpG islands depends on their transcriptional activity during differentiation and disease.

The human genome contains ∼30,000 CpG islands (CGIs). While CGIs associated with promoters nearly always remain unmethylated, many of the ∼9,000 CGIs lying within gene bodies become methylated during development and differentiation. Both promoter and intragenic CGIs may also become abnormally methylated as a result of genome rearrangements and in malignancy. The epigenetic mechanisms by which some CGIs become methylated but others, in the same cell, remain unmethylated in these situations are poorly understood. Analyzing specific loci and using a genome-wide analysis, we show that transcription running across CGIs, associated with specific chromatin modifications, is required for DNA methyltransferase 3B (DNMT3B)-mediated DNA methylation of many naturally occurring intragenic CGIs. Importantly, we also show that a subgroup of intragenic CGIs is not sensitive to this process of transcription-mediated methylation and that this correlates with their individual intrinsic capacity to initiate transcription in vivo. We propose a general model of how transcription could act as a primary determinant of the patterns of CGI methylation in normal development and differentiation, and in human disease.