TRANSMISSION // 2026.11.15.766802v1

A broader genomic map of maize breeding lines

Source published

New maize genome assemblies add structural and gene-level detail to resources for studying how genetic variation relates to crop traits.

AI-generated conceptual illustration for A broader genomic map of maize breeding lines
AI-generated conceptual illustration; not documentary evidence.

A single reference genome cannot capture all the genetic differences among maize lines. A new Genomes-to-Fields resource adds assemblies and annotations for 13 lines, creating material for future research without showing that any particular difference improves crop performance.

Why make more maize genomes?

Maize lines can differ in DNA sequence, gene content and chromosome structure. Such variation may matter for traits such as flowering time or yield, but a single reference genome cannot represent every version of the crop’s genome. The Genomes-to-Fields initiative connects maize genomic information with field-trait and environmental data. This study expands its genomic resources to help researchers investigate genetic diversity and, in future work, its relationship to traits.

The researchers present 15 genome assemblies representing 13 maize lines. The total exceeds the number of lines because two have partially phased assemblies, with two partitions retained for each. The set includes founders and testers from several Genomes-to-Fields evaluation cycles, along with additional breeding germplasm. The work describes genomic resources; it does not test whether particular DNA differences cause changes in crop performance.

How the genomes were built

The researchers used long-read sequencing, principally PacBio HiFi data, and incorporated Oxford Nanopore reads where appropriate. They assembled sequences using approaches adapted to individual lines, then assessed them using measures such as conserved-gene recovery, sequencing-read evidence and comparisons with the B73 reference.

For comparisons with B73, the team assigned and ordered many assembled sequences relative to that reference. Some scaffold joins were supported by reference order alone. The authors therefore caution that these scaffolds should be interpreted as chromosome assignments, not as uniformly continuous, independently resolved chromosome-scale assemblies.

The team also predicted gene models, assessed their completeness and likelihood of encoding proteins, compared genes with B73, annotated repetitive DNA and aligned genomes to identify large structural differences. These analyses map genomic features; they do not directly establish their effects on plant traits.

What the study found

The assemblies ranged from 2.19 to 2.38 billion DNA bases. Contig N50—a summary of the lengths of assembled sequence blocks—was 94.9–195.8 million bases for 10 primary assemblies. LH287 was more fragmented, with an N50 of 5.3 million bases. Across assemblies other than LH287, 95.9–99.4% of sequence was placed on chromosomes; for LH287, the figure was 90.8%. These placements partly rely on the reference-guided process described above. The assemblies recovered 97.9–99.0% of the conserved genes used to assess genome completeness.

The annotation pipeline predicted roughly 40,000 gene models per genotype. A computational assessment classified about 93.6–93.9% as likely protein-coding; this is not experimental confirmation that each gene is active or functional. Comparisons with B73 identified 27,606–30,517 one-to-one syntenic gene pairs per genotype. The authors note that a gene without a one-to-one match is not necessarily absent from B73: duplicate copies, changes in genomic location and annotation differences can affect the result.

A separate comparison using Liftoff matched a higher proportion of genes to B73 annotations, but it did not require one-to-one matches. As a result, multiple genes in a maize line could map to the same B73 gene. The two approaches therefore describe different kinds of correspondence rather than interchangeable counts.

Transposable-element annotations covered 83.6–88.6% of the nuclear genomes, with long-terminal-repeat retrotransposons the largest class. The analysis also identified three large chromosome inversions relative to B73. The largest, spanning 12.65 million bases on chromosome 3, was found in six of the 13 lines. The paper reports supporting evidence for these structural differences, though the strength and type of support varied among inversions and lines.

What remains uncertain

The resource broadens the maize genomes available for comparison, but it does not establish which differences influence agronomic traits or how effects might vary across environments. Those questions require linking genomic variation to measured traits and testing the links in further studies. The selected breeding lines do not represent all maize diversity, and assembly fragmentation or annotation choices can affect comparisons. Helitron detection also failed in three genotypes, so their total transposable-element estimates may be slightly low.

This is a bioRxiv preprint posted November 18, 2026, and it had not been certified by peer review at posting. Its findings and methods may change with review or later analysis. The paper lists raw sequencing reads at the European Nucleotide Archive, assemblies at MaizeGDB, and assembly code on GitHub. These resources allow researchers to examine and build on the work.

Automatically matched by shared topic keywords. These links do not imply that the sources agree or establish a consensus.