Image processing and supervised machine learning for retinal microglia characterization in senescence.

The process of senescence impairs the function of cells and can ultimately be a key factor in the development of disease. With an aging population, senescence-related diseases are increasing in prevalence. Therefore, understanding the mechanisms of cellular senescence within the central nervous system (CNS), including the retina, may yield new therapeutic pathways to slow or even prevent the development of neuro- and retinal degenerative diseases. One method of probing the changing functions of senescent retinal cells is to observe retinal microglial cells. Their morphological structure may change in response to their surrounding cellular environment. In this chapter, we show how microglial cells in the retina, which are implicated in aging and diseases of the CNS, can be identified, quantified, and classified into five distinct morphotypes using image processing and supervised machine learning algorithms. The process involves dissecting, staining, and mounting mouse retinas, before image capture via fluorescence microscopy. The resulting images can then be classified by morphotype using a support vector machine (SVM) we have recently described showing high accuracy. This SVM model uses shape metrics found to correspond with qualitative descriptions of the shape of each morphotype taken from existing literature. We encourage more objective and widespread use of methods of quantification such as this. We believe automatic delineation of the population of microglial cells in the retina, could potentially lead to their use as retinal imaging biomarkers for disease prediction in the future.

Projected future climatic forcing on the global distribution of vegetation types.

Most emissions scenarios suggest temperature and precipitation regimes will change dramatically across the globe over the next 500 years. These changes will have large impacts on the biosphere, with species forced to migrate to follow their preferred environmental conditions, therefore moving and fragmenting ecosystems. However, most projections of the impacts of climate change only reach 2100, limiting our understanding of the temporal scope of climate impacts, and potentially impeding suitable adaptive action. To address this data gap, we model future climate change every 20 years from 2000 to 2500 CE, under different CO2 emissions scenarios, using a general circulation model. We then apply a biome model to these modelled climate futures, to investigate shifts in climatic forcing on vegetation worldwide, the feasibility of the migration required to enact these modelled vegetation changes, and potential overlap with human land use based on modern-day anthromes. Under a business-as-usual scenario, up to 40% of terrestrial area is expected to be suited to a different biome by 2500. Cold-adapted biomes, particularly boreal forest and dry tundra, are predicted to experience the greatest losses of suitable area. Without mitigation, these changes could have severe consequences both for global biodiversity and the provision of ecosystem services. This article is part of the theme issue 'Ecological novelty and planetary stewardship: biodiversity dynamics in a transforming biosphere'.