Fig. 4

Improvement of TOP accuracy driven by robust omics data. A–C Identification rate of TOP increases when more omics traits are included in the model. For the Maize368 dataset, 17 agronomic traits (Agro), 88 transcriptomic traits (Exp), and 24 metabolic traits (Met) were sequentially added in the TOP model; For the Maize282 dataset, 21 agronomic traits, 144 transcriptomic traits from developing tissues (Exp1) and 182 transcriptomic traits from adult tissues (Exp2) were sequentially added in the TOP model; For Rice210 dataset, 4 agronomic traits (Agro), 46 transcriptomic traits (Exp), and 38 metabolic traits (Met) were included. All omics data with single-trait prediction accuracy less than 0.25 were excluded from the analyses. D–F Identification rate improvement due to filtering low-quality data. Before model training, traits with prediction accuracy (r) greater than 0.5 were considered; after training, traits with poor weights (w<0) were excluded from the model