Search

In plant breeding and genetics, predictive models traditionally rely on compact representations of high-dimensional data, often using methods like Principal Component Analysis (PCA) and, more recently, Autoencoders (AE). However, these methods do not separate genotype-specific and environment-specific features, limiting their ability to accurately predict traits influenced by both genetic and environmental factors. We hypothesize that disentangling these representations into genotype-specific and environment-specific components can enhance predictive models. To test this, we developed a compositional autoencoder (CAE) that decomposes high-dimensional data into distinct genotype-specific and environment-specific latent features. Our CAE framework employed a hierarchical architecture within an autoencoder to effectively separate these entangled latent features. Applied to a maize diversity panel dataset, the CAE demonstrated superior modeling of environmental influences and out-performs PCA (principal component analysis), PLSR (Partial Least square regression) and vanilla autoencoders by 7 times for ‘Days to Pollen’ trait and 10 times improved predictive performance for ‘Yield’. By disentangling latent features, the CAE provided a powerful tool for precision breeding and genetic research. This work has significantly enhanced trait prediction models, advancing agricultural and biological sciences.

Genome-wide DNA polymorphism analysis and molecular marker development are important for forward genetics research and DNA marker-assisted breeding. As an ideal model system for Panicoideae grasses and an important minor crop in East Asia, foxtail millet (Setaria italica) has a high-quality reference genome as well as large mutant libraries based on the “Yugu1” variety. However, there is still a lack of genetic and mutation mapping tools available for forward genetics research on S. italica. Here, we screened another S. italica genotype, “SSR41”, which is morphologically similar to, and readily cross-pollinates with, “Yugu1”. High-throughput resequencing of “SSR41” identified 1,102,064 reliable single nucleotide polymorphisms (SNPs) and 196,782 insertions/deletions (InDels) between the two genotypes, indicating that these two genotypes have high genetic diversity. Of the 8,361 high-quality InDels longer than 20 bp that were developed as molecular markers, 180 were validated with 91.5% accuracy. We used “SSR41” and these developed molecular markers to map the white leaf sheath gene SiWLS1. Further analyses showed that SiWLS1 encodes a chloroplast-localized protein that is involved in the regulation of chloroplast development in bundle sheath cells in the leaf sheath in S. italica and is related to sensitivity to heavy metals. Our study provides the methodology and an important resource for forward genetics research on Setaria.

Image-based high-throughput plant phenotyping in greenhouse has the potential to relieve the bottleneck currently presented by phenotypic scoring which limits the throughput of gene discovery and crop improvement efforts. Numerous studies have employed automated RGB imaging to characterize biomass and growth of agronomically important crops. The objective of this study was to investigate the utility of hyperspectral imaging for quantifying chemical properties of maize and soybean plants in vivo. These properties included leaf water content, as well as concentrations of macronutrients nitrogen (N), phosphorus (P), potassium (K), magnesium (Mg), calcium (Ca), and sulfur (S), and micronutrients sodium (Na), iron (Fe), manganese (Mn), boron (B), copper (Cu), and zinc (Zn). Hyperspectral images were collected from 60 maize and 60 soybean plants, each subjected to varying levels of either water deficit or nutrient limitation stress with the goal of creating a wide range of variation in the chemical properties of plant leaves. Plants were imaged on an automated conveyor belt system using a hyperspectral imager with a spectral range from 550 to 1,700 nm. Images were processed to extract reflectance spectrum from each plant and partial least squares regression models were developed to correlate spectral data with chemical data. Among all the chemical properties investigated, water content was predicted with the highest accuracy [R2 = 0.93 and RPD (Ratio of Performance to Deviation) = 3.8]. All macronutrients were also quantified satisfactorily (R2 from 0.69 to 0.92, RPD from 1.62 to 3.62), with N predicted best followed by P, K, and S. The micronutrients group showed lower prediction accuracy (R2 from 0.19 to 0.86, RPD from 1.09 to 2.69) than the macronutrient groups. Cu and Zn were best predicted, followed by Fe and Mn. Na and B were the only two properties that hyperspectral imaging was not able to quantify satisfactorily (R2 < 0.3 and RPD < 1.2). This study suggested the potential usefulness of hyperspectral imaging as a high-throughput phenotyping technology for plant chemical traits. Future research is needed to test the method more thoroughly by designing experiments to vary plant nutrients individually and cover more plant species, genotypes, and growth stages.

Identifying and characterizing alternative splicing (AS) enables our understanding of the biological role of transcript isoform diversity. This study describes the use of publicly available RNA-Seq data to identify and characterize the global diversity of AS isoforms in maize using the inbred lines B73 and Mo17, and a related species, sorghum. Identification and characterization of AS within maize tissues revealed that genes expressed in seed exhibit the largest differential AS relative to other tissues examined. Additionally, differences in AS between the two genotypes B73 and Mo17 are greatest within genes expressed in seed. We demonstrate that changes in the level of alternatively spliced transcripts (intron retention and exon skipping) do not solely reflect differences in total transcript abundance, and we present evidence that intron retention may act to fine-tune gene expression across seed development stages. Furthermore, we have identified temperature sensitive AS in maize and demonstrate that drought-induced changes in AS involve distinct sets of genes in reproductive and vegetative tissues. Examining our identified AS isoforms within B73 × Mo17 recombinant inbred lines (RILs) identified splicing QTL (sQTL). The 43.3% of cis-sQTL regulated junctions are actually identified as alternatively spliced junctions in our analysis, while 10 Mb windows on each side of 48.2% of trans-sQTLs overlap with splicing related genes. Using sorghum as an out-group enabled direct examination of loss or conservation of AS between homeologous genes representing the two subgenomes of maize. We identify several instances where AS isoforms that are conserved between one maize homeolog and its sorghum ortholog are absent from the second maize homeolog, suggesting that these AS isoforms may have been lost after the maize whole genome duplication event. This comprehensive analysis provides new insights into the complexity of AS in maize.

Conserved noncoding sequences (CNS) are islands of noncoding sequence that, like protein coding exons, show less divergence in sequence between related species than functionless DNA. Several of CNSs have been demonstrated experimentally to function as cis-regulatory regions. However, the specific functions of most CNSs remain unknown. Previous searchers for CNS in plants have either anchored on exons and only identified nearby sequences or required years of painstaking manual annotation. Here we present an open source tool that can accurately identify CNSs between any two related species with sequenced genomes, including both those immediately adjacent to exons and distal sequences separated by >12 KB of noncoding sequence. We have used this tool to characterize new motifs, associate CNSs with additional functions and identify previously undetected genes encoding RNA and protein in the genomes of five grass species. We provide a list of 15,363 orthologous CNSs conserved across all grasses tested. We were also able to identify regulatory sequences present in the common ancestor of grasses that have been lost in one or more extant grass lineages. Lists of orthologous gene pairs and associated CNSs are provided for reference inbred lines of arabidopsis, Japonica rice, foxtail millet, sorghum, brachypodium and maize.

Whole genome duplications, or tetraplodies, are an important source of increased gene content. Following whole genome duplication, duplicate copies of many genes are lost from the genome. This loss of genes is biased both in the classes of genes deleted and the subgenome from which they are lost. Many or all classes are genes preferentially retained as duplicate copies are engaged in dose sensitive protein-protein interactions, such that deletion of any one duplicate upsets the status quo of subunit concentrations, and presumably lowers fitness as a result. Transcription factors are also preferentially retained following every whole genome duplications studied. This has been explained as a consequence of protein-protein interactions, just as for other highly retained classes of genes. We show that the quantity of conserved non-coding sequences (CNSs) associated with genes predicts the likelyhood of their retention as duplicate pairs following whole genome duplication. As many CNSs likely represent binding sites for transcriptional regulators, we propose that the likelyhood of gene retention following tetraploidy may also be influenced by dose-sensitive protein-DNA interactions between the regulatory regions of CNS-rich genes -- nicknamed "bigfoot genes" – and the proteins that bind to them. Using grass genomes, we show that differential loss of CNSs from one member of a pair following the pregrass tetraploidy reduces its chance of retention in the subsequent maize-lineage tetraploidy.