• HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
Thursday, November 19, 2026
BIOENGINEER.ORG
No Result
View All Result
  • Login
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
No Result
View All Result
Bioengineer.org
No Result
View All Result
Home NEWS Science News Agriculture

Machine Learning and Network Modeling Identify Core Genes Shaping Brassica Leafy Heads

by
November 19, 2026
in Agriculture
Reading Time: 5 mins read
0
Machine Learning and Network Modeling Identify Core Genes Shaping Brassica Leafy Heads

Machine Learning and Network Modeling Identify Core Genes Shaping Brassica Leafy Heads

Share on FacebookShare on TwitterShare on LinkedinShare on RedditShare on Telegram

Computational models combining machine learning algorithms and gene regulatory network analysis have pinpointed key candidate genes and shared regulatory pathways that govern leafy head development in Chinese cabbage and cabbage. In a comparative investigation published in Molecular Horticulture, an international research team analyzed large-scale transcriptomic datasets from Brassica rapa and Brassica oleracea to address a longstanding biological question in horticultural genomics: how complex genetic networks coordinate the inward curling, compact architecture, and specialized morphological changes required to produce marketable leafy heads. While previous experimental studies had implicated individual hormonal pathways and leaf polarity determinants, the wider systems-level mechanisms coordinating this complex agronomical trait across divergent species had remained largely unresolved.

To build predictive models capable of discovering previously uncharacterized heading-related genes, the researchers compiled hundreds of public RNA-sequencing datasets from the National Center for Biotechnology Information SRA repository. After trimming raw reads with Trimmomatic to eliminate sequence artifacts and mapping them to reference genomes using STAR—aligning Brassica rapa reads to the Chiifu 3.5 assembly and Brassica oleracea reads to the JZS 2.0 assembly—the investigators quantified feature reads using Subread featureCounts and normalized the counts into log counts per million with edgeR. This standardized data pipeline yielded 634 normalized expression features for Brassica rapa, encompassing 444 features derived from heading morphotypes and 190 from non-heading varieties. In parallel, the Brassica oleracea dataset comprised 762 normalized expression features, split between 364 heading samples and 398 non-heading samples spanning diverse developmental conditions and botanical tissues.

A primary technical hurdle in applying supervised machine learning to plant developmental biology is the lack of confirmed negative training instances, as researchers rarely have verified ground truth regarding which genes play zero role in head development. To establish a robust ground truth set, the authors gathered 47 well-characterized heading genes from published functional literature to serve as the positive training class, designated CLASS H. These documented genes spanned vital developmental functions, including adaxial-abaxial leaf polarity specification, leaf curvature control, shoot apical meristem maintenance, small RNA activity, and plant hormone signaling. Rather than selecting negative genes entirely at random—which risks introducing severe expression-level disparities between classes—the team developed a systematic matching strategy. They binned genes by median expression values across all samples into groups of 100 and drew an equal cohort of 47 CLASS NH genes showing matching median expression profiles, thereby training the algorithms on true regulatory patterns rather than general expression abundance.

Using the randomForest R package, the researchers trained independent classification models for each species, parameterizing each ensemble with 500 decision trees and setting the split-variable parameter to 25 features for Brassica rapa and 27 features for Brassica oleracea. Rigorous five-fold cross-validation demonstrated strong predictive capability, yielding a mean Area Under the Receiver Operating Characteristic Curve of 0.87 plus or minus 0.08 for Brassica rapa and 0.85 plus or minus 0.04 for Brassica oleracea. Further evaluation through leave-one-out cross-validation across all 47 known heading genes confirmed statistical reliability, correctly predicting 83.0 percent of known genes in Brassica rapa and 76.6 percent in Brassica oleracea, both significantly exceeding chance expectations under hypergeometric testing. Well-established polarity and meristem regulators such as PHB.1, KNAT1, and YAB2.3 consistently achieved elevated prediction probabilities above 0.79 in both species.

Analysis of feature importance using Mean Decrease Accuracy and partial dependence profiling clarified how different plant organs contributed to the predictive power of the algorithms. When examining tissues shared by both species, the predictive models showed no significant differences in importance for leaf, root, shoot apical meristem, and seedling tissues, suggesting that core heading mechanisms rely on regulatory events in common vegetative structures. In contrast, reproductive structures such as flower and silique tissues showed significant divergence in feature importance scores between species, reflecting varying regulatory dynamics during their respective reproductive transitions. Partial dependence plots for the top root tissues revealed a consistent negative impact on heading prediction probabilities across nineteen out of twenty samples, corroborating biological expectations that heading-specific gene expression is concentrated within aerial vegetative tissues rather than subterranean roots.

When the finalized models were deployed across all annotated genes in both genomes, they flagged thousands of candidate genes—termed preGenes—exhibiting prediction probability scores exceeding 0.5. The models identified 5,201 preGenes in Brassica rapa and 8,101 preGenes in Brassica oleracea. Comparative analysis of syntenic orthologs located in corresponding genomic positions revealed 2,614 shared syntenic candidate pairs whose prediction scores exhibited a statistically significant positive correlation between the two species. Applying a stringent confidence threshold above 0.9 isolated an elite group of eleven unique high-confidence genes, with the transcription factor TCP15.3 reaching an identical high score of 0.91 in both models. Functional profiling through Gene Ontology enrichment demonstrated that these shared candidate pools were significantly enriched for biological processes central to head morphogenesis, including unidimensional cell growth, adaxial-abaxial pattern specification, meristem development regulation, and auxin transport.

To explore the spatial behavior of these predicted candidates, the authors evaluated the eleven top-scoring genes using an independent high-resolution transcriptomic dataset covering 24 spatially dissected tissues across the mature Chinese cabbage head. This examination revealed three distinct spatial expression profiles that aligned closely with known developmental functions. One cluster, containing TCP15.3, IMK3.1, FH10.1, LAX2.1, ATHB33.3, ZAR1.1, and BraA05g021110.3.5C, formed a steep expression gradient running from highest abundance in the innermost leaves to lower levels in outer leaves. A second group, comprising BUM.3, BLH9.3, and BraA07g000870.3.5C, displayed preferential expression restricted to the shoot apical meristem, consistent with roles in stem cell maintenance. The third pattern, represented by BraA07g023910.3.5C, showed consistently low expression across all surveyed leaf tissues, illustrating the diverse spatial deployment of candidate regulatory factors.

Recognizing that complex plant traits arise from coordinated networks rather than isolated loci, the researchers constructed genome-scale gene regulatory networks using GENIE3 to infer directed relationships between transcription factors and target genes. Filtering putative interactions with an edge weight threshold of 0.01 captured 332,429 regulatory links in Brassica rapa and 382,103 links in Brassica oleracea. Subsequent topological community partitioning via the iGraph package identified ten distinct sub-clusters in Brassica rapa and nine in Brassica oleracea. Intersecting the machine-learning candidates with these network topologies retained 18,792 candidate-specific interactions in Brassica rapa and 44,139 in Brassica oleracea, effectively mapping the predicted genes into tightly interconnected regulatory modules.

Comparative modular analysis identified two corresponding network modules that function as central regulatory engines for leafy head formation: cluster 8 in Brassica rapa and cluster 7 in Brassica oleracea. These two sub-networks shared the highest mutual Jaccard similarity index of 0.20 and harbored the greatest concentrations of established heading genes, with each cluster containing 27 documented ground truth loci, including YAB1, AS1, PHB, and KNAT family members. Furthermore, all twelve top-scoring candidate genes identified in preliminary rankings mapped exclusively within these two core modules. Gene Ontology analyses of both clusters revealed shared enrichment for primary morphogenetic programs, most notably adaxial-abaxial axis specification, epidermal tissue development, and auxin transport, demonstrating marked functional conservation despite evolutionary divergence.

To pinpoint the most influential regulatory hubs driving these key sub-networks, the investigators performed multi-metric network centrality screening using the CytoNCA Cytoscape plug-in. By calculating betweenness, closeness, degree, and eigenvector centralities and retaining only those genes that exceeded median values across all four metrics through three iterative selection cycles, they identified 50 core genes in Brassica rapa cluster 8 and 81 core genes in Brassica oleracea cluster 7. Conserved transcription factors emerging from this rigorous screen included ANT, GRF2, GRF3, and TCX3, alongside species-specific hubs such as GRF5 in cabbage. The authors emphasized that while these computational predictions provide high-confidence candidates for targeted Brassica breeding and functional dissection, empirical validation through transgenic transformation remains technically demanding due to genomic complexity, establishing a clear imperative for subsequent functional studies to verify these identified regulatory circuits in living plants.

Subject of Research: Computational identification of gene networks controlling leafy head formation in Brassica crops

Article Title: Integrative machine learning and gene regulatory network analysis identifies novel genes for leafy head formation in Brassica rapa and B. oleracea

Article References: Sun, W., Wu, J., Cai, X., Chen, Y., van Dijk, A. D. J., Wang, X., & Bonnema, G. (2026). Integrative machine learning and gene regulatory network analysis identifies novel genes for leafy head formation in Brassica rapa and B. oleracea. Molecular Horticulture, 6(1), Article 45. https://doi.org/10.1186/s43897-026-00230-1

Image Credits: AI Generated

DOI: 10.1186/s43897-026-00230-1

Keywords: Brassica rapa, Brassica oleracea, Chinese cabbage, cabbage, leafy head formation, machine learning, Random Forest, gene regulatory networks, transcriptomics, GENIE3, transcription factors, network centrality, syntenic genes, leaf polarity, auxin transport, plant development

News Source: Juliet Wilcox. (November 19, 2026). Machine Learning and Network Modeling Identify Core Genes Shaping Brassica Leafy Heads. Scienmag.

Tags: auxin transportBrassica oleraceaBrassica rapacabbageChinese cabbagegene regulatory networksGENIE3leaf polarityleafy head formationMachine Learningnetwork centralityplant developmentRandom Forestsyntenic genestranscription factorsTranscriptomics
Share12Tweet7Share2ShareShareShare1

Related Posts

Survey Finds Aflatoxins in 43.5 Percent of Kurdistan Cereal Samples

Survey Finds Aflatoxins in 43.5 Percent of Kurdistan Cereal Samples

November 19, 2026
Researchers Optimize Blueberry Shoot Regeneration and Transformation Systems Using Citrus Ruby1 Gene

Researchers Optimize Blueberry Shoot Regeneration and Transformation Systems Using Citrus Ruby1 Gene

November 19, 2026

Hyperspectral Leaf Petiole Reflectance Moderately Discriminates Latent Olive Pathogen Infections

November 19, 2026

Molecular Review Details How Cold Stress and Genetics Impact Fruit Quality

November 19, 2026

POPULAR NEWS

  • Updated ArrayAnalysis Platform Expands Interactive Analysis for Microarray and RNA-seq Data

    Updated ArrayAnalysis Platform Expands Interactive Analysis for Microarray and RNA-seq Data

    29 shares
    Share 12 Tweet 7
  • Stem Cells in Pituitary Tumors Show Multilineage Endocrine Potential

    29 shares
    Share 12 Tweet 7
  • Biotechnology Advances Expand Broodstock Management and Artificial Reproduction in Rainbow Trout

    29 shares
    Share 12 Tweet 7
  • Non-Vascular Ehlers-Danlos Surgery Carries High Complication Rates but No Life-Threatening Risks

    29 shares
    Share 12 Tweet 7

About

We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.

Follow us

Recent News

Degree of Nondensifiability Yields New Multivalued Fixed Point Theorems in Banach Algebras

Cancer Studies Omit Age and Sex Data on Drug Conjugate Nausea

Review Identifies Four Shared Circadian Biomarkers Linking Menopause and Alzheimer’s Disease

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 89 other subscribers
  • Contact Us

Bioengineer.org © Copyright 2023 All Rights Reserved.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Homepages
    • Home Page 1
    • Home Page 2
  • News
  • National
  • Business
  • Health
  • Lifestyle
  • Science

Bioengineer.org © Copyright 2023 All Rights Reserved.