Arabidopsis phenotyping through geometric morphometrics

Abstract

BackgroundRecently, great technical progress has been achieved in the field of plant phenotyping. High-throughput platforms and the development of improved algorithms for rosette image segmentation make it possible to extract shape and size parameters for genetic, physiological, and environmental studies on a large scale. The development of low-cost phenotyping platforms and freeware resources make it possible to widely expand phenotypic analysis tools for Arabidopsis. However, objective descriptors of shape parameters that could be used independently of the platform and segmentation software used are still lacking, and shape descriptions still rely on ad hoc or even contradictory descriptors, which could make comparisons difficult and perhaps inaccurate. Modern geometric morphometrics is a family of methods in quantitative biology proposed to be the main source of data and analytical tools in the emerging field of phenomics studies. Based on the location of landmarks (corresponding points) over imaged specimens and by combining geometry, multivariate analysis, and powerful statistical techniques, these tools offer the possibility to reproducibly and accurately account for shape variations among groups and measure them in shape distance units.ResultsHere, a particular scheme of landmark placement on Arabidopsis rosette images is proposed to study shape variation in viral infection processes. Shape differences between controls and infected plants are quantified throughout the infectious process and visualized. Quantitative comparisons between two unrelated ssRNA+ viruses are shown, and reproducibility issues are assessed.ConclusionsCombined with the newest automated platforms and plant segmentation procedures, geometric morphometric tools could boost phenotypic features extraction and processing in an objective, reproducible manner.

Morphological plasticity in a wild freshwater fish, Systomus sarana (Cyprinidae) from India: a glimpse through advanced morphometric toolkits 

“…The visualization of the body shape differences, associated with other groups of correlated morphological traits, allowed to obtain a clear diagnosis of sh morphology for each population [43,44]. Visualization tools might help in further study of the putative underlying mechanisms involved [45]. The result of the present study is in line with other studies based on truss analysis [4,[46][47][48][49] and geometric morphometrics [50,51] which have shown the sh species to have a distinctive morphology.…”

Walking on Kendall’s Shape Space: Understanding Shape Spaces and Their Coordinate Systems

“…Nevertheless, the tangent distances of specimens to the most outlying taxon, the common dolphin, were consistently slightly less than the partial Procrustes distances for the same pairwise comparisons, suggesting that there was some subtle effect of the distortion due to tangent projection for the extreme comparisons. In recent years, a number of studies have computed similar correlations between pairwise Procrustes and tangent distances among landmark configurations in diverse organisms and consistently found that these correlations were extremely close to 1.0 (Pretorius and Scholtz 2001;Polly 2002;Frost et al 2003;Fontaneto et al 2004;Lockwood et al 2004;McNulty 2004;Viscosi and Cardini 2011;Bai et al 2012;De Meulemeester et al 2012;Renner 2012;Neustupa 2013;Siver et al 2013;Ullmann et al 2017;Manacorda and Asurmedi 2018). Rohlf (1999) commented that he was not aware of any reports of biological datasets where the tangent projection did not provide a satisfactory approximation (with the exception of inadvertent inclusion of reflected specimens in the data).…”

ML-morph: A Fast, Accurate and General Approach for Automated Detection and Landmarking of Biological Structures in Images

“…These discoveries were only possible through the development of a technological infrastructure that allowed us to acquire genomic information at a large scale (Schuster, 2007). While many researchers have argued that phenomics -large scale phenotyping -will bring about a similar revolution in biology, most approaches for collecting high-throughput phenotypic data developed so far are system-specific and difficult to generalize (see Kristensen et al, 2008;Falkingham, 2012;Boyer et al, 2011;Hsiang et al, 2018;Manacorda and Asurmendi, 2018), with the notable exception of sub-cellular phenotypes (e.g., Clish, 2015). As a consequence, for most study systems, phenotyping methods remain low-throughput and can only be applied on a small scale (Houle, Govindaraju, and Omholt, 2010).…”

“…A promising way to collect high-throughput phenotypic data is to automate landmark data collection using computer vision techniques (e.g., Manacorda and Asurmendi, 2018). Automated landmarking has become the gold standard in human facial landmarking for both biomedicine (Porto et al, 2019) and, more notoriously, social networking websites and software developed for mobile phones (Kazemi and Sullivan, 2014), but its application in geometric morphometrics has remained restricted (Manacorda and Asurmendi, 2018;Vandaele et al, 2018).…”

“…A promising way to collect high-throughput phenotypic data is to automate landmark data collection using computer vision techniques (e.g., Manacorda and Asurmendi, 2018). Automated landmarking has become the gold standard in human facial landmarking for both biomedicine (Porto et al, 2019) and, more notoriously, social networking websites and software developed for mobile phones (Kazemi and Sullivan, 2014), but its application in geometric morphometrics has remained restricted (Manacorda and Asurmendi, 2018;Vandaele et al, 2018). However, the explosion in machine learning algorithms for computer vision represents an important technological leap, which lays the foundation for the development of general methods of high-throughput high-dimensional morphometrics (Kazemi and Sullivan, 2014;He et al, 2017;Voulodimos et al, 2018).…”