From genotype to phenotype in Arabidopsis thaliana: in-silico genome interpretation predicts 288 phenotypes from sequencing data

Abstract

In many cases, the unprecedented availability of data provided by high-throughput sequencing has shifted the bottleneck from a data availability issue to a data interpretation issue, thus delaying the promised breakthroughs in genetics and precision medicine, for what concerns Human genetics, and phenotype prediction to improve plant adaptation to climate change and resistance to bioagressors, for what concerns plant sciences. In this paper, we propose a novel Genome Interpretation paradigm, which aims at directly modeling the genotype-to-phenotype relationship, and we focus on A. thaliana since it is the best studied model organism in plant genetics. Our model, called Galiana, is the first end-to-end Neural Network (NN) approach following the genomes in/phenotypes out paradigm and it is trained to predict 288 real-valued Arabidopsis thaliana phenotypes from Whole Genome sequencing data. We show that 75 of these phenotypes are predicted with a Pearson correlation ≥0.4, and are mostly related to flowering traits. We show that our end-to-end NN approach achieves better performances and larger phenotype coverage than models predicting single phenotypes from the GWAS-derived known associated genes. Galiana is also fully interpretable, thanks to the Saliency Maps gradient-based approaches. We followed this interpretation approach to identify 36 novel genes that are likely to be associated with flowering traits, finding evidence for 6 of them in the existing literature.

Figure 1.

The architecture of Galiana. The (17, N = 27 655) tensorial representation of each genome is used as input. Each 17-dimensional vector representing the gene i is processed by the G module. The iterated applications of the G module produces the (1, 27 655) representation of the mutational load on each gene, which is used as input for the fully connected module, which consists of a FF NN with two layers with 50 neurons each. Finally, the last layer implements the multi-task regression over the 288 real-valued phenotypes.

Figure 2.

Scatter plots showing the prediction results for 16 of the 75 significantly predicted phenotypes.

Figure 3.

Plots visualizing the predicted correlations for four selected phenotypes while considering only AT samples located in Sweden, Italy, Spain and Russia (represented by the columns, from left to right). This shows that Galiana predicts also intra-nation phenotype dynamics and not only among AT samples belonging to different countries.

Figure 4.

Visual comparison between our multi-phenotypic predictions (Galiana) and the single-phenotype models (GWAS_NN_i) based on the known associated genes retrieved from (40). Galiana outperforms the GWAS-based NN 77% of the times on the 35 phenotypes on which the GWAS_NN was applicable.