Biology

Conventional machine learning approaches for hyperspectral crop trait estimation usually optimize preprocessing, feature selection, sample partitioning, and regression as separate steps. This can overlook interdependencies among steps and limit model performance and robustness. A two-tier grid search framework was developed to systematically evaluate 576 combinations of data splitting, spectral preprocessing, band selection, and regression methods for maize leaf SPAD estimation. Maize leaf SPAD values were estimated from hyperspectral images of 208 samples collected under well-watered and drought-stressed conditions, with SPAD-502 readings as ground truth. The best-performing pipeline was D2 + LARS + LGBM under KS splitting. Specifically, to ensure robustness and prevent data leakage, wavelength selection was performed exclusively on the training set prior to data splitting, and the selected wavelengths were subsequently applied to the test set. After hyperparameter tuning, it achieved test-set R² = 0.7160, RMSE = 2.2504, and MAE = 1.8946. Spatial inversion revealed within-leaf SPAD heterogeneity under drought stress. GLCM-based texture and leaf coverage analyses were used to evaluate changes in leaf spatial patterns and plant morphology under drought stress. Overall, this study provides a simple pipeline-level optimization strategy for hyperspectral crop trait modeling, reducing reliance on manual trial-and-error, and offering a potentially useful framework for precision agriculture and remote sensing-based phenotyping applications.

Satellite DNAs (satDNAs) are highly repetitive sequences that contribute to genome organization, chromosome structure and evolution. Although studies in several Drosophila species have advanced our understanding of satDNA biology, most species remain unexplored, leaving their evolutionary dynamics and origins poorly understood in a broader context. Here, we characterized the satellitome of the spotted-wing drosophila, Drosophila suzukii (Diptera: Drosophilidae), an economically important pest, integrating computational analyses with cytogenetic mapping. Remarkably, the species presented a high amount of C-positive blocks, mainly on Chr-2. Collectively, the satDNAs identified using RepeatExplorer2/TAREAN (Galaxy Version 2.3.12.1) represented approximately 3.3% of the genome, highlighting striking differences in chromosomal distribution and abundance. While two satDNAs are amplified in C-positive regions, particularly on Chr-2, contributing to the remarkable expansion of C-positive blocks in this chromosome, other elements have a more localized distribution. Despite distinct abundance and chromosomal distribution patterns, the high level of sequence homogeneity within families suggests recent and/or sustained amplification and homogenization. The satDNA family DsuzSat04-163 originates from the tandemization of the central part of the Helitron-1N1_Dsuz transposable element (TE). Moreover, we characterized two TEs forming dimer arrays, highlighting TEs as potential sources of and contributors to the diversification of satDNA sequences in D. suzukii. Finally, comparative chromosomal and assembly-based mapping revealed that discrete satDNA loci are well represented, whereas cytologically defined centromeric regions and pericentromeric repeats remain poorly resolved, likely due to the collapse or incomplete reconstruction of highly repetitive C-positive regions. Together, these findings uncover the hidden complexity of repetitive DNA in D. suzukii and highlight the value of satellitome characterization for understanding genome organization and improving its genomic representation.

Brown cotton is the most dominant type of naturally colored cotton (NCC), yet breeding it to combine desirable color with high fiber quality remains challenging. The previously developed FLP (pFbL2A-Lc-2A-GhPAP1D) transgenic cotton showed visually uneven reddish-brown pigmentation and defective secondary cell wall (SCW) development, which ultimately compromises fiber quality and yield. To address these issues, two RNA interference (RNAi) strategies coupled with conventional crossing were adopted to modify FLP fibers: one aimed at enhancing reddish pigmentation, and the other at improving fiber quality. Unexpectedly, SCW-specific silencing of proanthocyanidin (PA) synthesis genes GhANR, GhLAR1 and GhLAR2 driven by the pFbL2A promoter failed to enhance red coloration and instead resulted in lighter fibers compared with the FLP9 parent, showing no significant difference from the null segregants. In contrast, pEXPA2-mediated repression of the Lc-2A-GhPAP1D fusion gene during the fiber transition stage effectively rescued the defective fiber development, substantially improving fiber quality and lint percentage relative to both the FLP9 parent and the null segregants, but at the expense of pigmentation. This study establishes and compares two gene-silencing systems for colored cotton, offering valuable insights into the trade-off between pigmentation and fiber quality for future breeding programs.