RÉSUMÉ
Disruption of retinal vasculature is linked to various diseases, including diabetic retinopathy and macular degeneration, leading to vision loss. We present here a novel algorithmic approach that generates highly realistic digital models of human retinal blood vessels, based on established biophysical principles, including fully-connected arterial and venous trees with a single inlet and outlet. This approach, using physics-informed generative adversarial networks (PI-GAN), enables the segmentation and reconstruction of blood vessel networks with no human input and which out-performs human labelling. Segmentation of DRIVE and STARE retina photograph datasets provided near state-of-the-art vessel segmentation, with training on only a small (n = 100) simulated dataset. Our findings highlight the potential of PI-GAN for accurate retinal vasculature characterization, with implications for improving early disease detection, monitoring disease progression, and improving patient care.
Sujet(s)
Algorithmes , Apprentissage profond , Rétine , Vaisseaux rétiniens , Humains , Vaisseaux rétiniens/imagerie diagnostique , Rétine/imagerie diagnostique , Traitement d'image par ordinateur/méthodes , Rétinopathie diabétique/diagnostic , Dégénérescence maculaire/anatomopathologieRÉSUMÉ
BACKGROUND: Carcinogenesis is driven by interactions between genetic mutations and the local tumor microenvironment. Recent research has identified hundreds of cancer driver genes; however, these studies often include a mixture of different molecular subtypes and ecological niches and ignore the impact of the immune system. RESULTS: In this study, we compare the landscape of driver genes in tumors that escaped the immune system (escape +) versus those that did not (escape -). We analyze 9896 primary tumors from The Cancer Genome Atlas using the ratio of non-synonymous to synonymous mutations (dN/dS) and find 85 driver genes, including 27 and 16 novel genes, in escape - and escape + tumors, respectively. The dN/dS of driver genes in immune escaped tumors is significantly lower and closer to neutrality than in non-escaped tumors, suggesting selection buffering in driver genes fueled by immune escape. Additionally, we find that immune evasion leads to more mutated sites, a diverse array of mutational signatures and is linked to tumor prognosis. CONCLUSIONS: Our findings highlight the need for improved patient stratification to identify new therapeutic targets for cancer treatment.
Sujet(s)
Mutation , Tumeurs , Échappement de la tumeur à la surveillance immunitaire , Humains , Tumeurs/génétique , Tumeurs/immunologie , Échappement de la tumeur à la surveillance immunitaire/génétique , Échappement immunitaire/génétique , Évolution moléculaire , Microenvironnement tumoral/génétiqueRÉSUMÉ
Though the earliest stages of oncogenesis, post initiation, are not well understood, it is generally appreciated that a successful transition from a collection of dysregulated cells to an aggressive tumour requires complex ecological interactions between cancer cells and their environment. One key component of tumorigenesis is immune evasion. To investigate the interplay amongst the ecological behaviour of mutualism and immune evasion, we used a computational simulation framework. Sensitivity analyses of the growth of a virtual tumour implemented as a 2D-hexagonal lattice model suggests tumour survival depends on the interplay between growth rates, mutualism and immune evasion. In 60% of simulations, cancer clones with low growth rates, but exhibiting mutualism were able to evade the immune system and continue progressing suggesting that tumours with equivalent growth rates and no mutualism are more likely to be eliminated than tumours with mutualism. Tumours with faster growth rates showed a lower dependence upon mutualism for progression. Geostatistical analysis showed decreased spatial heterogeneity over time for polyclonal tumours with a high division rate. Overall, these results suggest that in slow growing tumours, mutualism is critical for early tumorigenesis.