Alex Whan

Key message: Latent environmental effects of genotype by environment interactions could be predicted from observed environmental covariates. Predictions into the wider target population of environments revealed greater insights. Wheat is grown across a diverse range of environments in Australia with contrasting environmental constraints. Targeted breeding to optimise genotypes in target environments is hindered by large and ubiquitous genotype by environment interactions (GEI). Common GEI in multi-environment trial experiments, which sample the target population of environments, can be efficiently modelled using latent environmental effects from factor analytic mixed models. However, generalised prediction into the full target population of environments is difficult without a clear link to observed environmental covariates (ECs) that are defined from high-resolution weather and soil data. Here, we used a large wheat multi-environment trial dataset and demonstrated that latent environmental effects can be associated with and predicted from observed ECs. We found GEI-based environment classes could be defined by combinations of key ECs. Prediction of main and latent effects in a wider set of environments covering the full TPE across the Australian grain belt over 13 years revealed the complex trends of environmental effects and GEI over regional scales demonstrating high year-to-year variability. Regional environment types often shifted year-to-year. Cross-validation of forward genomic prediction into untested year environments demonstrated that increased accuracy is possible if estimated genetic effects are also accurate and ECs of new environments are known. These findings may guide Australian wheat breeders to better target specifically adapted material to mega-environments defined by static GEI while also considering broad adaptability and non-static GEI resulting from year-to-year variability.

Genome annotation is essential for understanding the functional elements within genomes. While automated methods are indispensable for processing large-scale genomic data, they often face challenges in accurately predicting gene structures and functions. Consequently, manual curation by domain experts remains crucial for validating and refining these predictions. These combined outcomes from automated tools and manual curation highlight the importance of integrating human expertise with artificial intelligence (AI) capabilities to improve both the accuracy and efficiency of genome annotation. However, the manual curation process is inherently labor-intensive and time-consuming, making it difficult to scale for large datasets. To address these challenges, we propose a conceptual framework, Human-AI Collaborative Genome Annotation (HAICoGA), that leverages the synergistic partnership between humans and AI to enhance human capabilities and accelerate the genome annotation process. Additionally, we explore the potential of integrating large language models into this framework to support and augment specific tasks. Finally, we discuss emerging challenges and outline open research questions to guide further exploration in this area.

Identifying the genetic architecture of complex traits requires access to populations with sufficient genetic diversity and recombination. Multiparent Advanced Generation InterCross (MAGIC) populations are a powerful resource due to their balanced population structure, allelic diversity, and enhanced recombination. However, implementing a MAGIC population in complex polyploids such as wheat is challenging, as wheat harbors many introgressions, inversions, and other genetic factors that interfere with linkage mapping. By utilizing a comprehensive crossing strategy, additional rounds of mixing, and novel genotype calling approaches, we developed a bread wheat 8-parent MAGIC population of over 3,000 genotyped recombinant inbred lines derived from 2,151 distinct crosses. This effort resulted in a dense genetic map covering the complete genome. Further rounds of intercrossing led to increased recombination in inbred lines, as expected. We identified structural variation highlighted by segregation distortion, along with epistatic allelic interactions between specific founders. We report on a novel and effective resource for genomic and trait exploration in hexaploid wheat, capable of detecting small genetic effects and epistatic interactions due to the high level of recombination and large number of lines. The interactions and genetic effects identified provide a basis for ongoing research to understand the basis of allelic frequencies across the genome, particularly where economically important loci are involved.

Root architecture is key in determining how effective plants are at intercepting and absorbing nutrients and water. Previously, the wheat (Triticum aestivum) cultivars Spica and Maringa were shown to have contrasting root morphologies. These cultivars were crossed to generate an F_6:1 population of recombinant inbred lines (RILs) which was genotyped using a 90 K single nucleotide polymorphisms (SNP) chip. A total of 227 recombinant inbred lines (RILs) were grown in soil for 21 days in replicated trials under controlled conditions. At harvest, the plants were scored for seven root traits and two shoot traits. An average of 7.5 quantitative trait loci (QTL) were associated with each trait and, for each of these, physical locations of the flanking markers were identified using the Chinese Spring reference genome. We also compiled a list of genes from wheat and other monocotyledons that have previously been associated with root growth and morphology to determine their physical locations on the Chinese Spring reference genome. This allowed us to determine whether the QTL discovered in our study encompassed genes previously associated with root morphology in wheat or other monocotyledons. Furthermore, it allowed us to establish if the QTL were co-located with the QTL identified from previously published studies. The parental lines together with the genetic markers generated here will enable specific root traits to be introgressed into elite wheat lines. Moreover, the comprehensive list of genes associated with root development, and their physical locations, will be a useful resource for researchers investigating the genetics of root morphology in cereals.

Motivation: Genetic map construction is a foundational step in the analysis of structured experimental populations. For markers that hybridize to several genetically similar locations, or where several alleles are present (such as in multiparental populations), current methods often discard the marker or incorrectly call the genotypes. These errors result in information loss, or incorrect genotypes that can corrupt map construction. Results: We present a new approach for simultaneously performing genetic map construction and marker calling. Our new approach allows the calling of a larger number of markers, a larger number of unique alleles per marker and the correct use of markers which hybridize to multiple genetically similar locations. We demonstrate our new approach using simulations, a biparental wheat population and an eight-parent population of spring bread wheat. Applying our method to the eight-parent population increased the number of mapped markers by 71%. We show that the new genetic map allows the investigation of synteny in ways that were not previously possible in that dataset. Availability and implementation: The method described in this article has been incorporated into R package mpMap2. It is available from CRAN and also from https://github.com/rohan-shah/mpMap2. Supplementary information: Supplementary data are available at Bioinformatics online.

Background: Identifying the genetic architecture of complex traits requires access to populations with sufficient genetic diversity and recombination. Multi-parent Advanced Generation InterCross (MAGIC) populations are a powerful resource due to their balanced population structure, allelic diversity and enhanced recombination. However, implementing a MAGIC population in complex polyploids such as wheat is not trivial, as wheat harbours many introgressions, inversions and other genetic factors that interfere with linkage mapping. Results: By utilising a comprehensive crossing strategy, additional rounds of mixing and novel genotype calling approaches, we developed a bread wheat eight parent MAGIC population made up of more than 3000 fully genotyped recombinant inbred lines derived from 2151 distinct crosses, and achieved a dense genetic map covering the complete genome. Further rounds of inter-crossing led to increased recombination in inbred lines, as expected. The comprehensive and novel approaches taken in the development and analysis of this population provide a platform for genetic discovery in bread wheat. We identify previously unreported structural variation highlighted by segregation distortion, along with the identification of epistatic allelic interactions between specific founders. We demonstrate the ability to conduct high resolution QTL mapping using the number of recombination events as a trait, and identify several significant QTLs explaining greater than 50% of the variance. Conclusions: We report on a novel and effective resource for genomic and trait exploration in hexaploid wheat, that can be used to detect small genetic effects and epistatic interactions due to the high level of recombination and large number of lines. The interactions and genetic effects identified provide a basis for ongoing research to understand the basis of allelic frequencies across the genome, particularly where economically important loci are involved.

Motivation: Major advances have been made in the assembly of complex genomes such as wheat. Remaining challenges include linking these new resources to legacy research, as well as lowering the bar to entry for members of the research community who may lack specific bioinformatics skills. Results: Pretzel aims to solve these problems by providing an interactive, online environment for data visualisation and analysis which, when loaded with appropriately curated data, can enable researchers with no bioinformatics training to exploit the latest genomic resources. We demonstrate that Pretzel can be used to answer common questions asked by researchers as well as for advanced curation of pseudomolecule structure. Availability: Pretzel is implemented in JavaScript and is freely available under a GPLv3 license online at: https://github.com/plantinformatics/pretzel . Contact: gabriel.keeble-gagnere@ecodev.vic.gov.au

Realising the potential of large genetic resources requires the ability to perform genotyping efficiently. In populations with tens or hundreds of thousands of SNP markers, it is infeasible to assign genotypes manually; this must be done automatically. This problem is generally solved by the application of mixture models (Xiao et al., 2007; Teo et al., 2007). We describe a marker calling method able to correctly identify marker heterozygotes, in large populations generated according to some experimental design; e.g. the Collaborative Cross (Threadgill and Churchill, 2012) or MAGIC (Huang et al., 2015). Identifying marker heterozygotes is of substantial value in these populations, which are substantially inbred with a small amount of residual heterozygosity. If genetic lines can be identified which are heterozygous at a QTL locus, these lines can be used to develop a follow-up population suitable for fine-mapping of that QTL. However, as the amount of residual heterozygosity is low, these lines cannot be identified by an unsupervised method. We describe a model-based clustering method for identifying marker heterozygotes in significantly inbred populations. Like Xiao et al. (2007) and Teo et al. (2007), our method is also based on mixture models. Unlike those methods, we include a constraint that the heterozygote cluster lies between the two marker homozygote clusters. This constraint accurately represents the structure of the heterozygote cluster in the case of data generated by the Illuimina Infinium genotyping platform. The proposed model-based clustering algorithm was motivated by an eight-parent MAGIC population in spring bread wheat, genotyped using the Illumina Infinium genotyping platform, and we demonstrate its use in that population. The proposed method is implemented in the R package magicCalling.

Late maturity α-amylase (LMA) and pre-harvest sprouting (PHS) are both recognized as environmentally induced grain quality defects resulting from abnormally high levels of α-amylase. LMA is a more recently identified quality issue that is now receiving increasing attention worldwide and whose prevalence is now seen as impeding the development of superior quality wheat varieties. LMA is a genetic defect present in specific wheat genotypes and is characterized by elevated levels of the high pI TaAMY1 α-amylase, triggered by environmental stress during wheat grain development. TaAMY1 remains present in the aleurone through the harvest, lowering Falling Number (FN) at receival, causing a down-grading of the grain, often to feed grade, thus reducing the farmers’ income. This downgrading is based on the assumption within the grain industry that, as for PHS, a low FN represents poor quality grain. Consequently any wheat line possessing low FN or high α-amylase levels is automatically considered a poor bread wheat despite there being no published evidence to date, to show that LMA is detrimental to end product quality. To evaluate the validity of this assumption a comprehensive evaluation of baking properties was performed from LMA prone lines using a subset of tall non-Rht lines from a multi-parent advanced generation inter-cross (MAGIC) wheat population grown at three different sites. LMA levels were determined along with quality parameters including end product functionality such as oven spring, bread loaf volume and weight, slice area and brightness, gas cell number and crumb firmness. No consistent or significant phenotypic correlation was found between LMA related FN and any of the quality traits. This manuscript provides for the first time, compelling evidence that LMA has limited impact on bread baking end product functionality.

In glasshouse studies we have previously shown that endosperm-specific RNAi suppression of the primary starch phosphorylation enzyme, Glucan, Water Dikinase (GWD) leads to enhanced early vigor, greater leaf biomass, and increases in both head size and yield. To confirm these affects in a field setting, trials were conducted in three Australian environments. Field results were consistent with those in the glasshouse for increased flag leaf area and rachis nodes. However, there was also a decrease in tiller number and consequently a decrease in yield for one event at two sites. These findings provide potentially important information on plant vigor enhancement and highlight the challenges of transferring the modification of complex traits from single plants in controlled environments to the field.