Statistics of seed phenotypic traits
The descriptive statistics for each trait assessed, under both control and accelerated aging conditions, revealed significant insights into seed germination and seedling growth (Supplementary Table S1). The frequency distribution curve exhibited a pattern closely resembling a normal distribution for germination rate index after accelerated aging (GRIAA), relative germination rate index after accelerated aging (GRIAAR), mean germination time (days) after accelerated aging (MGTAA), relative mean germination time after accelerated aging (MGTAAR), Shoot length (cm) after accelerated aging (SLAA), relative shoot length after accelerated aging (SLAAR) and relative seedling dry weight after accelerated aging (SDWAAR) indicating their polygenic nature (Fig. 1). AA had a detrimental effect on seed performance, leading to decreased seed germination percentage, germination rates, seedling growth and vigor across all the traits over the respective control (Fig. 2). These observations indicated that seeds subjected to AA experienced impaired metabolic activity, potentially due to oxidative stress and loss of membrane integrity during the aging process. The reduced seedling growth and lower vigor indices observed under AA further reflect the cumulative impact of aging on seed health, affecting both physiological and morphological aspects.
Frequency distribution of seed traits of the finger millet GWAS panel, the X-axis represents the trait values, and the Y-axis represents frequency. The traits are organized as follows: (a) GC, GAA, GAAR; (b) GIC, GIAA, GIAAR; (c) GRIC, GRIAA, GRIAAR; (d) MGTC, MGTAA, MGTAAR; (e) RLC, RLAA, RLAAR; (f) SLC, SLAA, SLAAR; (g) SVI1C, SVI1AA, SVI1AAR; (h) SDWC, SDWAA, SDWAAR; (i) SVI2C, SVI2AA, SVI2AAR.
Boxplot comparison of seed traits, the X-axis represents the trait names, and the Y-axis represents their corresponding values. (a) control vs. accelerated aging (b) relative values of traits after accelerated aging.
Initial germination of seeds (GC) ranged from 85.33 to 97.53%, with a mean of 93.57%. In comparison, germination after accelerated aging (GAA) showed a broader range of 23.26–95.52%, averaging 80.50%. Relative germination after accelerated aging (GAAR) varied significantly from 25.81 to 103.69%, highlighting the impact of accelerated aging on seed viability at different tolerance levels of genotypes. Germination index (GIC) ranged between 619.11 and 777.26, with an average of 721.01. Germination index after accelerated aging (GIAA) dropped considerably, ranging from 54.08 to 582.01, with a mean of 397.80. This represents nearly a two-fold reduction in germination potential, as reflected in relative germination index after accelerated aging (GIAAR), which averaged 55.13%, with some genotypes experiencing a significant decline (Supplementary Table S1).
Mean germination time (MGTC) averaged 1.29 days, with a range of 1.05 to 1.86 days. Mean germination time significantly increased to an average of 4.12 days, with a range of Germination index after accelerated aging 2.38 to 7.30 days after aging treatment. MGTAAR ranged from 229.29 to 472.63%, indicating that a longer period was needed for germination after the aging process. Root length (RLC) averaged 5.93 cm, with values ranging from 3.99 to 8.09 cm. After aging, root length after accelerated aging (RLAA) decreased to an average of 4.37 cm, with a range of 1.04 to 7.35 cm. Relative root length after accelerated aging (RLAAR) varied broadly, spanning from 18.94 to 130.46%, indicating a substantial decrease in radicle growth post-aging, with some genotypes showing a significant contraction. Shoot length (SLC) averaged 2.83 cm, with a range of 2.03 to 3.47 cm. Post-aging, SLAA dropped to an average of 1.99 cm, with values ranging from 0.44 to 2.87 cm, indicating a substantial reduction in shoot length due to aging.
Seedling vigor index-1 (SVI1C) averaged 818.68, with a range of 597.58 to 1066.41. Following aging, Seedling vigor index-1 (SVI1AA) decreased to an average of 516.17, ranging from 56.06 to 913.01. Relative seedling vigor index-1 after accelerated aging (SVI1AAR) ranged from 7.76 to 115.42%, with an average of 64.32%, indicating a considerable decline in seedling vigour. Seedling dry weight (SDWC) averaged 2.07 mg, ranging from 1.40 to 3.08 mg. Seedling dry weight after accelerated aging (SDWAA) slightly increased to an average of 2.16 mg, with a range of 0.58 to 3.20 mg. Relative seedling dry weight after accelerated aging (SDWAAR) varied from 34.64 to 141.75%, with some genotypes showing a substantial increase in dry weight despite the aging treatment. Seedling vigor index-2 (SVI2C) averaged 193.68, with values ranging from 127.33 to 281.24. Seedling vigor index-2 after aging, (SVI2AA) decreased to an average of 174.52, with a range of 24.97 to 256.89.
These results highlight significant variability in seed germination indices and seedling vigor traits under both control and AA conditions. The germination parameters, such as GC, GAA and GAAR, demonstrated substantial differences, reflecting the impact of AA on seed viability, which will help to predict the seed longevity potential of genotypes. The germination indices GIC, GIAA, GRIC, GRIAA and respective relative values (GIAAR and GRIAAR) revealed the speed of germination. Higher relative values for MGTAAR indicated the delay in the germination process of the genotype that reflects the poor vigor status of aged seeds. Similarly, seedling growth metrics, including root and shoot lengths (RLC and SLC) and their relative measures post-aging (RLAAR and SLAAR), measured the tolerance levels of growth under aging treatment.
The variance components showed significant differences for all the traits under control and AA conditions (Table 1). GC had a heritability of 61.22%, while GAA and GAAR showed heritability of 90.03% and 86.80%, respectively. Traits related to germination indices viz., GIC, GIAA, and GIAAR had a heritability of 78.33%, 93.06%, and 91.73%, respectively. Mean germination time (MGTC, MGTAA, MGTAAR), root and shoot lengths (RLC, SLC) and seedling vigor indices (SVI1C, SVI2C) had heritability ranging from 80.63 to 95.08%. These results explained the significant genetic basis of seed germination and seedling growth traits, indicating the potential for genetic enhancement of seed vigor and longevity.
GIAA showed significant positive correlations with GAA (0.86), GIAAR (0.97), and GRIAA (0.98) (Fig. 3). On the other hand, MGTAA exhibited a negative correlation with traits such as SVI1AA (-0.60) and SVI2AA (-0.28). A significant negative correlation between GAA and MGTAA (-0.55) indicates that seeds with higher germination rates after aging stress tend to germinate early. Similarly, the significant negative correlation between GIAA (-0.87) and GRIAA (-0.86) with MGTAA suggests that early germination is associated with better overall seed performance. These relationships provide valuable insights for breeding strategies focused on improving seed viability, early germination and robust seedling growth under aging conditions. The correlations among multiple traits highlight potential genetic linkages and pathways influencing seed development and viability, which could be critical for breeding programs to enhance finger millet resilience and productivity.
Correlation heatmap of seed traits in finger millet indicating trait relationships.
Statistics of SNP calling
A total of 5,63,270 SNPs were identified using the GBS approach from a panel of 221 genotypes. After eliminating SNPs with over 10% missing values, 1,04,907 SNPs were retained. These SNPs were then filtered by removing more than 10% heterozygosity and a minor allele frequency (MAF) below 3%, resulting in the retention of 11,832 high-quality SNPs for the genome-wide association study (Fig. 4).
Flow diagram describing the in-silico pipeline used for SNP filtering.
The SNP density across the chromosomes was analysed and presented in the heatmap (Fig. 5) (Supplementary Table S2). Chromosome 5 A had the highest SNP density (1054), followed by chromosome 5B (1101) and chromosome 6 A (1014). Chromosome 3B had the lowest SNPs (164), while scaffold regions of the genome contained 50 SNPs.
SNP density across the chromosomes of finger millet illustrating 11,832 SNPs in 1 Mb window size. The X-axis represents chromosome positions (Mb), and the Y-axis shows chromosome number, with red indicating the highest density and green the lowest.
Linkage disequilibrium (LD)
LD was measured across 11,832 SNP markers among 221 finger millet accessions, revealing that the LD block size at LD50 decay was 3.39 Mb (Fig. 6). SNPs within this distance are considered to belong to the same inheritance block.
The linkage disequilibrium (LD) decay plot among 221 diverse finger millet genotypes in GWAS panel, featuring 11,832 SNPs. The X-axis represents the genomic distance in Mb, while the Y-axis displays the r² values. The red line marks the threshold where linkage LD decayed to 50% of its maximum value.
Principle Component Analysis (PCA)
PCA revealed a relatively uniform distribution of genotypes across regions, with some overlap between populations. African and Asian genotypes are grouped separately, suggesting the presence of subpopulation structure. (Fig. 7a). The PCA based on races showed uniform distribution and wider spread across the first three principal components (Fig. 7b).
Principal components for 221 genotypes illustrating population structure: (a) diverse geographical origins, (b) different races.
Marker trait associations (MTAs)
The GAPIT model selection feature, which integrates both the kinship matrix and PCA, was employed to identify the associations with the traits in a panel of 221 genotypes using 11,832 high-quality SNPs.
A total of 491 MTAs was discovered with a significance -log10(p) value greater than 3. Among traits, SVI2AA had the highest number of significant MTAs (32) and RLAAR had the fewest (4). The SNP associations for each trait under different treatments, along with their p-values are presented in Supplementary Table S3. MTAs were further filtered using a Bonferroni threshold of -log10(p) greater than 5.37, which narrowed down to 54 SNPs across 13 chromosomes. Table 2 provides detailed information on those Bonferroni-corrected MTAs. Furthermore, Manhattan plots displaying the MTAs and Q-Q plots showing the observed versus expected associations of various seed traits, adjusted for population structure, are presented in Figs. 8, 9 and 10.
FM_SNP_10872 was identified as the most significant marker associated with the trait SDWC. The pleiotropic quantitative trait loci (QTL), FM_SNP_9478 on chromosome 7B, exhibited associations with multiple traits, including GAA, GAAR, GIAA and GIAAR. GAAR, SDWAAR and SVI2AAR shared a common association with FM_SNP_235 on chromosome 2A. Another pleiotropic locus, FM_SNP_9403 explained 41.94% of phenotypic variance (PV) for SVI2C, 19.79% for SDWC and 17.99% for SDWAA. Comprehensive statistics for the significant SNPs identified are provided in Table 2.
Seed germination percentage
Nine significant markers were identified for QTLs associated with seed germination percentage, with three linked to GAA and six to GAAR. FM_SNP_11315 on chromosome 6B was observed in both treated seeds and relative germination after accelerated aging. For GAA, three SNPs on chromosomes 5B, 6B and 7B were identified, with FM_SNP_9478 on 7B being highly significant and explaining 31.52% of PV. For GAAR, six SNPs on chromosomes 1 A, 1B, 2A, 5B, 6B and 7B were identified, with FM_SNP_235 on 2A and FM_SNP_9478 on 7B explaining 28.36% and 18.63% of PV, respectively.
Manhattan plot of genome-wide SNPs of (a) GC, GAA, GAAR (b) GIC, GIAA, GIAAR (c) GRIC, GRIAA, GRIAAR. The X-axis shows genomic positions, and the Y-axis shows − log10(p-values). The grey horizontal line indicates the Bonferroni correction threshold. The corresponding Q–Q plot displays observed versus expected − log10(p-values).
Germination indices
The loci, FM_SNP_9478 and FM_SNP_9582 on chromosome 7B were observed in both treated seeds and the relative germination index after AA treatment. The locus FM_SNP_3796 on chromosome 8A in treated seeds showed the strongest association, explaining 31.12% of PV. For GRIC, only one SNP, FM_SNP_3303 on chromosome 6A, was linked, explaining 46.07% of PV. No significant associations were detected for GRIAA and GRIAAR.
Mean germination time
Significant SNPs associated with mean germination time were identified for control and relative mean germination time under AA treatment. For MGTC, FM_SNP_3303 and FM_SNP_10858 located on chromosomes 6A and 6B explained a PV of 31.86% and 30.14%, respectively. Similarly, FM_SNP_8722 on chromosome 9B showed a significant association and explained 12.03% of PV. For MGTAAR, FM_SNP_9852 on chromosome 5B explained 32.17% of PV.
Manhattan plot of genome-wide SNPs of (a) MGTC, MGTAA, MGTAAR (b) RLC, RLAA, RLAAR (c) SDWC, SDWAA, SDWAAR. X-axis showing genomic positions and the Y-axis showing − log10(p-values). The grey horizontal line indicates the Bonferroni threshold. The corresponding Q–Q plot displays observed versus expected − log10(p-values).
Root and shoot length
FM_SNP_8168 on chromosome 7B was associated with RLC, explaining 27.08% of PV. While, FM_SNP_3851 on chromosome 8 A showed a strong association, accounting for 39.58% of PV. No significant SNP associations were found for RLAA and RLAAR. FM_SNP_3696 on chromosome 3A was identified for SLC, accounted for 32.48% of PV. No significant SNPs were observed for SLAA and SLAAR.
Seedling dry weight
Among all the traits, seed germination percentage revealed the highest number of MTAs, with eight SNPs identified for SDWC and three for SDWAA. Additionally, four SNPs were identified for SDWAAR. The locus FM_SNP_9403 on chromosome 1B was consistently detected in control and treated seeds after the Bonferroni correction. For SDWC, FM_SNP_9403 on chromosome 1B and FM_SNP_10872 on chromosome 6B showed strong statistical significance explaining 19.79% and 21.56% of PV, respectively. Similarly, SDWAA showed significant associations with FM_SNP_4948 on chromosome 3A and FM_SNP_483 on chromosome 5A. These SNPs explained a PV of 37.48% and 30.75%, respectively. FM_SNP_1517 on chromosome 1A as a significant SNP with 70.19% of PV was identified for SDWAAR, highlighting its substantial role in influencing seedling response under AA condition over the control.
Manhattan plot of genome-wide SNPs of (a) SLC, SLAA, SLAAR (b) SVI1C, SVI1AA, SVI1AAR (c) SVI2C, SVI2AA, SVI2AAR. X-axis showing genomic positions and the Y-axis showing − log10(p-values). The grey horizontal line indicates the Bonferroni correction threshold. The corresponding Q-Q plot displays observed versus expected − log10(p-values).
Seedling vigor indices
GWAS model identified three significant SNPs associated with SVI1C, but did not find any significant associations for SVI1AA and SVI1AAR. FM_SNP_6503, located on chromosome 1A, explained 34.31% of PV, while FM_SNP_3851 on chromosome 8A explained 25.99% of PV. Significant SNPs associated with SVI2C, SVI2AA and SVI2AAR were primarily identified under control conditions. FM_SNP_9403 on chromosome 1B explained 41.94% of PV, whereas FM_SNP_8521 on chromosome 9B accounted for 23.58% of PV in SVI2C highlighting their significant influence on the trait. In contrast, SVI2AAR was linked to FM_SNP_235 on chromosome 2A and FM_SNP_9246 on chromosome 6B, explained 46.17% and 10.97% of PV, respectively, under rapid aging.
Several pleiotropic loci, namely, FM_SNP_9403, FM_SNP_9478, FM_SNP_3851, FM_SNP_9852 and FM_SNP_235, were detected for multiple traits. The SNP, FM_SNP_9478 on chromosome 7B was significantly associated with GAA, GAAR, GIAA and GIAAR, suggesting a common genetic basis in predicting seed longevity. The correlation study further confirmed that these four traits were significantly and positively associated. FM_SNP_7145 on chromosome 5B and the SNP FM_SNP_11315 on chromosome 6B were both linked to both GAA and GAAR, indicating the genetic relationship between these traits. These findings explained a strong correlation between the seed traits and the presence of common SNPs, suggesting potential pleiotropy or closely linked genes influencing multiple phenotypic characteristics in finger millet. These hotspots could serve as valuable markers for improving multiple traits through marker-assisted selection.
The high heritability values for traits viz., GIAA (93.06%), GAAR (86.80%) and GAA (90.03%) indicated a strong genetic control of these seed longevity characteristics. Notably, the SNPs associated with these high heritability traits also explained a significant portion of the phenotypic variation for seed storability potential. SNP FM_SNP_9478 on chromosome 7B accounted for 31.52% of the variation in GAA and 19.19% in GIAA. Similarly, FM_SNP_7145 on chromosome 5B explained 22.67% of the variation in GAA and 9.89% in GAAR, while FM_SNP_11315 on chromosome 6B explained 15.15% of the PV in GAA and 7.54% in GAAR. These findings suggested that the traits with high heritability are also those explaining a high degree of phenotypic variance for seed longevity.
In-silico comparative genomics
Significant SNPs associated with seed quality traits were functionally annotated across different monocot species viz., rice (Oryza sativa), maize (Zea mays), foxtail millet (Setaria italica), sorghum (Sorghum bicolor) and switchgrass (Panicum virgatum) to determine their roles in seed development, metabolism and adaptation. Considering the LD span of 3.39 Mb, genes within a 2 Mb region around the significant SNPs in finger millet were annotated. Among the SNPs identified as significantly associated through GWAS, annotations were obtained for 24 SNPs linked to crucial seed traits across various monocot species (Supplementary Table S4). FM_SNP_6503 and FM_SNP_6470 in sorghum were associated with the light-mediated development protein DET1 and probable protein phosphatase 2C3, respectively. FM_SNP_1517 and FM_SNP_9403 were associated with expansin-A2 and piezo-type mechanosensitive ion channels, indicating their roles in root mechanotransduction. Respiratory burst oxidase homolog protein F in sorghum for FM_SNP_51 and auxin transport protein BIG-like in switchgrass for FM_SNP_6154 were identified. DP-dependent glyceraldehyde-3-phosphate dehydrogenase for FM_SNP_3881, beta-amylase for FM_SNP_9582, and auxin transport protein BIG for FM_SNP_6154 were identified as other important genes. This highlights their potential roles in mitigating accelerated aging effects through regulatory mechanisms in seed development.
GO analysis highlighted enrichment in key biological processes such as embryo development ending in seed dormancy and ribosome biogenesis (Fig. 11a). KEGG pathway analysis identified significant pathways such as ribosome, endocytosis, phagosome and proteosome, emphasizing their importance in maintaining seed vigor and longevity, especially under conditions of accelerated aging (Fig. 11b). This comprehensive analysis enhances our understanding of the genetic basis of seed vigor and longevity traits in finger millet and reveals conserved molecular mechanisms shared with related cereal crops.
(a) GO analysis illustrating finger millet’s molecular landscape: Cellular components, biological processes and molecular functions. The Y-axis shows the fold enrichment and the X-axis shows highly enriched GO categories. (b) Dot plot visualization of KEGG pathway enrichment: Fold enrichment (x-axis) vs. pathway name (y-axis). Dot size represents gene count and color indicates p-value threshold.
