Quantitative evaluation of nonlinear methods for population structure visualization and inference.

Population structure (also called genetic structure and population stratification) is the presence of a systematic difference in allele frequencies between subpopulations in a population as a result of nonrandom mating between individuals. It can be informative of genetic ancestry, and in the context of medical genetics, it is an important confounding variable in genome-wide association studies. Recently, many nonlinear dimensionality reduction techniques have been proposed for the population structure visualization task. However, an objective comparison of these techniques has so far been missing from the literature. In this article, we discuss the previously proposed nonlinear techniques and some of their potential weaknesses. We then propose a novel quantitative evaluation methodology for comparing these nonlinear techniques, based on populations for which pedigree is known a priori either through artificial selection or simulation. Based on this evaluation metric, we find graph-based algorithms such as t-SNE and UMAP to be superior to principal component analysis, while neural network-based methods fall behind.

Deep neural networks for genomic prediction do not estimate marker effects.

Genomic prediction is a promising technology for advancing both plant and animal breeding, with many different prediction models evaluated in the literature. It has been suggested that the ability of powerful nonlinear models, such as deep neural networks, to capture complex epistatic effects between markers offers advantages for genomic prediction. However, these methods tend not to outperform classical linear methods, leaving it an open question why this capacity to model nonlinear effects does not seem to result in better predictive capability. In this work, we propose the theory that, because of a previously described principle called shortcut learning, deep neural networks tend to base their predictions on overall genetic relatedness rather than on the effects of particular markers such as epistatic effects. Using several datasets of crop plants [lentil (Lens culinaris Medik.), wheat (Triticum aestivum L.), and Brassica carinata A. Braun], we demonstrate the network's indifference to the values of the markers by showing that the same network, provided with only the locations of matches between markers for two individuals, is able to perform prediction to the same level of accuracy.

Latent Space Phenotyping: Automatic Image-Based Phenotyping for Treatment Studies.

Association mapping studies have enabled researchers to identify candidate loci for many important environmental tolerance factors, including agronomically relevant tolerance traits in plants. However, traditional genome-by-environment studies such as these require a phenotyping pipeline which is capable of accurately measuring stress responses, typically in an automated high-throughput context using image processing. In this work, we present Latent Space Phenotyping (LSP), a novel phenotyping method which is able to automatically detect and quantify response-to-treatment directly from images. We demonstrate example applications using data from an interspecific cross of the model C4 grass Setaria, a diversity panel of sorghum (S. bicolor), and the founder panel for a nested association mapping population of canola (Brassica napus L.). Using two synthetically generated image datasets, we then show that LSP is able to successfully recover the simulated QTL in both simple and complex synthetic imagery. We propose LSP as an alternative to traditional image analysis methods for phenotyping, enabling the phenotyping of arbitrary and potentially complex response traits without the need for engineering-complicated image-processing pipelines.

The use of plant models in deep learning: an application to leaf counting in rosette plants.

Deep learning presents many opportunities for image-based plant phenotyping. Here we consider the capability of deep convolutional neural networks to perform the leaf counting task. Deep learning techniques typically require large and diverse datasets to learn generalizable models without providing a priori an engineered algorithm for performing the task. This requirement is challenging, however, for applications in the plant phenotyping field, where available datasets are often small and the costs associated with generating new data are high. In this work we propose a new method for augmenting plant phenotyping datasets using rendered images of synthetic plants. We demonstrate that the use of high-quality 3D synthetic plants to augment a dataset can improve performance on the leaf counting task. We also show that the ability of the model to generate an arbitrary distribution of phenotypes mitigates the problem of dataset shift when training and testing on different datasets. Finally, we show that real and synthetic plants are significantly interchangeable when training a neural network on the leaf counting task.

Deep Plant Phenomics: A Deep Learning Platform for Complex Plant Phenotyping Tasks.

Plant phenomics has received increasing interest in recent years in an attempt to bridge the genotype-to-phenotype knowledge gap. There is a need for expanded high-throughput phenotyping capabilities to keep up with an increasing amount of data from high-dimensional imaging sensors and the desire to measure more complex phenotypic traits (Knecht et al., 2016). In this paper, we introduce an open-source deep learning tool called Deep Plant Phenomics. This tool provides pre-trained neural networks for several common plant phenotyping tasks, as well as an easy platform that can be used by plant scientists to train models for their own phenotyping applications. We report performance results on three plant phenotyping benchmarks from the literature, including state of the art performance on leaf counting, as well as the first published results for the mutant classification and age regression tasks for Arabidopsis thaliana.

Corrigendum: Deep Plant Phenomics: A Deep Learning Platform for Complex Plant Phenotyping Tasks.

[This corrects the article on p. 1190 in vol. 8, PMID: 28736569.].

A randomized controlled trial of attention modification for social anxiety disorder.

Social Anxiety Disorder (SAD) models implicate social threat cue vigilance (i.e., attentional biases) in symptom development and maintenance. A modified dot-probe protocol has been shown to reduce SAD symptoms, in some but not all studies, presumably by modifying an attentional bias. The current randomized controlled trial was designed to replicate and extend such research. Participants included treatment-seeking adults (n = 108; 58% women) who met diagnostic criteria for SAD. Participants were randomly assigned to a standard (i.e., control) or modified (i.e., active) dot-probe protocol condition and to participate in-lab or at home. The protocol involved twice-weekly 15-min sessions, for 4 weeks, with questionnaires completed at baseline, post-treatment, 4-month follow-up, and 8-month follow-up. Symptom reports were assessed with repeated measures mixed hierarchical modeling. There was a main effect of time from baseline to post-treatment wherein social anxiety symptoms declined significantly (p < .05) but depression and trait anxiety did not (p > .05). There were no significant interactions based on condition or participation location (ps > .05). Reductions were maintained at 8-month follow-up. Symptom reductions were not correlated with threat biases as indexed by the dot-probe task. The modified and standard protocol both produced significant sustained symptom reductions, whether administered in-lab or at home. There were no robust differences based on protocol type. As such, the mechanisms for benefits associated with modified dot-probe protocols warrant additional research.