Journal: Genome Biology

Article Title: FSBLUP: a novel strategy of fusion similarity matrix construction via optimally integrating intermediate omics data to enhance genomic prediction

doi: 10.1186/s13059-026-03931-4

Figure Lengend Snippet: Schematic overview of the fusion similarity best linear unbiased prediction framework. A Fusion similarity matrix construction integrating multi-source genomic data through block-matrix covariance propagation. Matrix panels: genomic similarity matrix (G, green), pedigree matrix (A, gray), and intermediate omics-derived matrix (M, blue). The fusion matrix (center) combines these layers via \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{C}=\alpha \mathrm{M}+\beta \mathrm{G}+\left(1-\alpha -\beta \right)\mathrm{A}$$\end{document} C = α M + β G + 1 - α - β A , where optimally weighted parameters ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha ,\beta$$\end{document} α , β ) balance contributions of each data layer to the final phenotypic outcomes. Missing multi-omics information for unmeasured individuals (Group 1) is inferred through covariance propagation. B Two-stage parameter optimization: (i) Grid search identifies high-accuracy regions; (ii) Adaptive bisection iteratively refines \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha ,\upbeta$$\end{document} α , β through contracting search windows (red points), guided by accuracy landscapes (contours). Convergence occurs after max iterations or when prediction accuracy gain is less than a pre-set threshold, e.g. , \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${10}^{-4}$$\end{document} 10 - 4 (dashed threshold)

Article Snippet: For individuals with complete multi-omics data (center block, Group 4), genomic ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathbf{G}$$\end{document} G ), intermediate omics ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathbf{M}$$\end{document} M ), and pedigree ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathbf{A}$$\end{document} A ) similarities are fused via \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathbf{C}=\alpha \mathbf{M}+\beta \mathbf{G}+\left(1-\alpha -\beta \right)\mathbf{A}$$\end{document} C = α M + β G + - α - β A , with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha /\beta$$\end{document} α / β represents the contributions of each data layer to the final phenotype.

Techniques: Blocking Assay, Derivative Assay, Biomarker Discovery

Journal: Genome Biology

Article Title: FSBLUP: a novel strategy of fusion similarity matrix construction via optimally integrating intermediate omics data to enhance genomic prediction

doi: 10.1186/s13059-026-03931-4

Figure Lengend Snippet: Comparison of prediction performances of various methods for wheat yield. A Nine methods for predicting grain yields of 588 bread wheat lines. Genetic value prediction accuracy was estimated using two-fold cross-validation, and 50% of the yield values were masked during model training. Genetic value prediction accuracy was estimated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rho }_{g}=co{r}_{g}\left(\widehat{u},y\right)\sqrt{{h}^{2}\left(\widehat{u}\right)}$$\end{document} ρ g = c o r g u ^ , y h 2 u ^ because hyperspectral data and actual yields were collected on the same plots. The bars represent average estimates (± standard deviation) over 20 replicate cross-validation runs for each method . Details of each model are presented in the section. B Phenotypic correlation (black lines) and estimates of genetic correlation (red lines) between each hyperspectral wavelength measured on each of the 9 flight dates with final grain yield. Genetic correlations were estimated with the GBLUP method using complete data. Vegetative growth (VEG), heading (HEAD), and grain filling (GF) represent different developmental growth stages. C Comparison of computing times (in seconds) of various methods. The y-axis represents the computing time on a log10 scale. Computing performance tests were performed on a Red Hat Enterprise Linux server with 2.60 GHz Intel(R) Xeon(R) Ice Lake 6348 CPU, and 256 GB memory. D Prediction with hyperspectral reflectance data in different developmental growth stages for Grain Yields of 588 bread wheat lines

Article Snippet: For individuals with complete multi-omics data (center block, Group 4), genomic ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathbf{G}$$\end{document} G ), intermediate omics ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathbf{M}$$\end{document} M ), and pedigree ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathbf{A}$$\end{document} A ) similarities are fused via \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathbf{C}=\alpha \mathbf{M}+\beta \mathbf{G}+\left(1-\alpha -\beta \right)\mathbf{A}$$\end{document} C = α M + β G + - α - β A , with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha /\beta$$\end{document} α / β represents the contributions of each data layer to the final phenotype.

Techniques: Comparison, Biomarker Discovery, Standard Deviation