Genome-wide Marker-based dissection of genetic variability for yield and yield components, and multi-trait selection in Kersting’s groundnut (Macrotyloma geocarpum)

doi:10.21203/rs.3.rs-4831288/v1

Download PDF

Article

Genome-wide Marker-based dissection of genetic variability for yield and yield components, and multi-trait selection in Kersting’s groundnut (Macrotyloma geocarpum)

https://doi.org/10.21203/rs.3.rs-4831288/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Understanding genetic parameters and trait relationships is essential for effective breeding programs. This study evaluated fifteen yield components in 81 kersting’s groundnut accessions from Benin and Burkina Faso using additive and dominant marker-based relationship matrices and mixed effect models. Structural equation modeling was used to assess trait association patterns, while genotype by yield-trait biplot and multi-trait genotype-ideotype distance index identified superior genotypes. Significant accession and environmental effects were observed for most traits. Broad-sense heritability was moderate for yield (H² = 0.39) and high (H² > 0.60) for other traits. Seed width, length, thickness, 100-seed weight, emergence rate, and days to 50% flowering exhibited moderate narrow-sense heritability (h² = 0.33–0.58), indicating additive gene action amenable to selection. High genetic gains were observed for emergence rate (84.09%), yield/plant (48.00%), and grain yield (44.19%), while low gains were found for number of pods/plant (0.32%), grain filling duration (2.60%) and pod width (3.72%). Grain yield exhibited significant positive associations with yield/plant, number of pods/plant, seeds/pod, 100-seed weight, pod harvest efficiency, and number of branches. These traits can guide indirect selection strategies for enhancing grain yield. Seventeen superior accessions with distinct strengths and weaknesses were identified as candidate parents for kersting’s groundnut breeding programs.

Biological sciences/Plant sciences

Biological sciences/Plant sciences/Plant breeding

variance components

heritability

additive effect

orphan crop

structural equation modeling

multi-trait selection

Kersting’s groundnut (Macrotyloma geocarpum (Harms) Maréchal & Baudet) is a priority crop for promoting food and nutrition security and alleviating poverty among smallholder farmers in Benin ¹. Its edible grains are rich sources of protein, essential amino acids, and minerals including Na (23.4 mg/100g), Ca (183.3 mg/100g), Mg (1.5 mg/100g), Zn (25.2 mg/100g) ² and P (345.5 mg/100g) ³. The grains also contain vitamins such as A (29.6 mg/100g), B1 (2.8 mg/100g) and B2 (0.2 mg/100g) ², as well as carbohydrates (56.5–59.4 g/100d dw) ⁴. However, the current production falls short of meeting the growing demand for kersting’s groundnut. The low productivity of the crop, around 500 kg ha^-1, is far below its genetic potential of 1.5 tons ha^{-1 5}, due to biotic and abiotic stresses ⁶ and the lack of improved varieties ⁷. To date, there is no formal kersting’s groundnut seed system and farmers still rely on unimproved landraces ⁸.

To design an efficient breeding program, it is necessary to have exhaustive information on genetic variability, heritability, and genetic advance ^9–11. Heritability estimates determine the extent of genetic control in the expression of a trait of interest and therefore serve as a measure of the selection efficiency based on that trait ^12,13. High narrow sense heritability, suggests a rapid response to selection ¹⁴. Akohoue, et al. ¹⁵ recently estimated broad sense heritability for yield and related traits in kersting’s groundnut and found high values for hundred seed weight (H² = 0.71), days to 50% flowering (H² = 0.86) and days to maturity (H² = 0.87). However, the narrow sense heritability of these traits has not been assessed. Therefore, to achieve quick genetic gain, it is important to dissect the genetic variability of grain yield and key yield-contributing traits to estimate narrow sense heritability. Partitioning genetic variance components can be done by making crosses and evaluating progenies, but also using molecular markers ^16,17. The latter approach which is largely used in human genetics ^18–20, has only a few reports in crops, despite its relevance in partitioning complex quantitative traits ²¹.

Various agronomic traits are considered in breeding programs, with yield being always the main concern ²². However, yield is a complex quantitative trait that can be deconstructed into several less complex, yield-contributing component traits that are more amenable to selection ^23–25. As such, breeding programs often target these yield component traits, as selecting for optimal combinations of yield components enables more efficient prediction and selection for overall yield improvement ²². To select for optimal trait combinations, plant breeders use selection indexes in which individual traits are weighted based on their relative economic importance for yield ^22,26,27. Despite the efficiency of use exhibited by selection indexes, some rules must be followed while using them. For instance, Ibrahim, et al. ²⁸ observed that gain from selection for any trait is likely to decrease as additional traits are included in the selection index, ringing the bell of caution of objective and accurate choice of the traits to be included in the selection index. Therefore, the identification of traits with high contribution to grain yields is an important step towards the selection of high-yielding kersting’s groundnut genotypes. In that process, path coefficient analysis is a widely used method to assess yield-contributing traits in several crops ^29–31, including kersting’s groundnut ¹⁵. One of the limitations of path coefficient analysis is its reliance on the assumption that the posited relationships between variables are correct ³². This can be problematic because, in reality, the true relationships between complex traits like yield and hypothetical yield components are often unknown or uncertain ^33,34. Basing path analysis on flawed or oversimplified assumptions about these relationships can lead to biased or misleading results and interpretations ³⁴. Structural equation modeling (SEM) is an alternative to path coefficient analysis that is better suited for dissecting yield components, as it enables accounting for bidirectional relationships and testing multiple competing models to empirically determine the causal relationships between variables ³². A major strength of SEM over path analysis is that it does not assume a single correct model, but rather uses data-driven model comparisons and goodness-of-fit assessments to determine the most likely causal networks. SEM has been used to analyze yield components in barley (Hordeum vulgare L.) ³⁵, wheat (Triticum aestivum L.) ³⁶, ryegrass (Lolium perenne L.) ³⁷, and maize ³⁸. This allowed reveal association trends that would not be observed in simple univariate or bivariate analyses ³². The power of SEM was thus used to examine the causal relationships between kersting’s groundnut grain yield and yield-contributing traits.

There are several selection indexes used to select superior plant genotypes ³⁹. Popular indices used by plant breeders include the Smith-Hazel ^40,41, Pesek and Baker ⁴², and Williams ⁴³ indexes. The main challenge with these indexes is the difficulty to define realistic economic weights for the traits of interest, limiting their utilization for selecting the best plant genotypes ⁴⁴. To overcome this limitation, Olivoto and Nardino ⁴⁵ have proposed the multi-trait genotype-ideotype distance index (MGIDI), which provides a more efficient and accurate selection of superior genotypes based on desired or undesired characteristics of the crop. This index has been used to select superior genotypes in crops such as wheat ^46,47, barley ²², and maize ^48,49.

The objectives of this research were: (i) to estimate genetic parameters for grain yield and yield components; (ii); to assess the relationships between grain yield and yield components; and (iii) to identify promising accessions that can be used in breeding programs.

Mean performance and phenotypic variation

The likelihood-ratio test (LRT) revealed significant (P < 0.05) variations among accessions for all traits, except number of pods per plant (NPP) and number of seeds per pod (NSP) (Table 1). A significant (P < 0.05) replication effect was found for five traits including pod length (PDL), pod width (PDW), seed weight (SDW), seed thickness (SDT), and NSP. Block effect was significant (P < 0.05) for eight traits including number of branches (NBR), days to 50% flowering (DTF), PDL, NSP, grain filling duration (GFD), 100-seed weight (HSW), yield per plant (YPP), and grain yield (GY). Most traits had coefficients of variation below 20% indicating a low dispersion of the experimental data (Table 2). Nevertheless, higher coefficients of variation were observed for emergence rate (EMR; 36.93%), NBR (33.27%), NPP (44.78%), YPP (47.80%), and GY (46.43%) indicating a high dispersion for those traits.

Table 1

Likelihood-ratio test (LRT) of random (accession and block within replication) and Wald statistics of fixed effect (replication) for yield related traits in kersting's groundnut
Trait	Code	Accession	Block [Rep]	Rep
Emergence rate (%)	EMR	59.64 ^***	2.10 ^ns	3.07 ^ns
Number of branches	NBR	14.87 ^***	88.33 ^***	1.62 ^ns
Days to 50% flowering	DTF	6.69 ^**	23.01^***	2.53 ^ns
Number of pods per plant	NPP	0.00 ^ns	0.98^ns	0.04 ^ns
Pod length (mm)	PDL	16.09^***	28.41^***	18.56 ^***
Pod width (mm)	PDW	6.39^*	0.11^ns	7.51 ^**
Pod harvest efficiency (%)	PHE	5.65^*	0.00 ^ns	0.58 ^ns
Number of seeds per pod	NSP	1.52 ^ns	5.40 ^*	23.69 _***
Grain filling duration (days)	GFD	30.27 ^***	98.97 ^***	0.73 ^ns
Seed length (mm)	SDL	49.08 ^***	1.50 ^ns	2.30 ^ns
Seed width (mm)	SDW	33.89 ^***	1.73 ^ns	4.85 ^*
Seed thickness (mm)	SDT	5.24 ^*	0.00 ^ns	20.06 ^***
100-seed weight (g)	HSW	27.74 ^***	6.50 ^*	2.69 ^ns
Yield per plant (g plant^− 1)	YPP	5.95 ^*	15.42 ^***	0.27 ^ns
Grain yield (kg ha^− 1)	GY	4.73 ^*	13.17 ^***	0.29 ^ns
* p < 0.05, p < 0.01, * p < 0.001, ns: not significant at 5% probability level, Rep: Replication.

Table 2

Descriptive statistics of kersting’s groundnut yield related traits
Trait	Code	Mean ± s.e.	Min	Max	CV (%)
Emergence rate (%)	EMR	0.44 ± 0.01	0.07	0.87	36.93
Number of branches	NBR	9.45 ± 0.20	2.30	17.90	33.27
Days to 50% flowering	DTF	48.82 ± 0.12	46.00	54.00	3.84
Number of pods per plant	NPP	91.21 ± 2.64	12.00	213.56	44.78
Pod length (mm)	PDL	12.40 ± 0.10	9.71	16.72	12.53
Pod width (mm)	PDW	7.80 ± 0.02	7.00	8.82	3.69
Pod harvest efficiency (%)	PHE	0.77 ± 0.01	0.48	0.96	12.91
Number of seeds per pod	NSP	1.29 ± 0.01	1.00	1.73	13.51
Grain filling duration (days)	GFD	57.26 ± 0.15	51.00	63.00	4.06
Seed length (mm)	SDL	8.20 ± 0.02	7.38	9.26	4.25
Seed width (mm)	SDW	5.68 ± 0.02	4.97	6.29	4.36
Seed thickness (mm)	SDT	4.26 ± 0.01	3.77	4.83	4.60
100-seed weight (g)	HSW	12.89 ± 0.09	9.25	16.35	10.29
Yield per plant (g plant^− 1)	YPP	11.48 ± 0.35	1.62	25.51	47.80
Grain yield (kg ha^− 1)	GY	506.40 ± 15.21	115.23	1043.46	46.43
CV: Coefficient of variation, s.e.: Standard error

Mean performance for all traits is presented in Table 2. A wide range was observed for GY (115.23–1043.46 kg ha^-1), YPP (1.62–25.51 g plant^-1), NPP (12–213.56 pods plant^-1), PHE (0.48–0.96) and EMR (0.07–0.87%) indicating substantial variability for these traits. In contrast, a smaller range of means was found for NSP (1.00–1.73 seeds pod^-1), SDT (3.77–4.83 mm), and SDW (4.74 − 6.42 mm) indicating less variability for these traits in the studied germplasm.

[Please insert Tables and around here]

Data quality, variance components and genetic parameters

The experimental accuracy, variance components, and genetic parameters estimates among the test kersting’s groundnut accessions are presented in Table 3. Experimental accuracy estimates were used to assess the data quality. Experimental accuracy was high (r_gg> 0.70) for all traits except NPP (r_gg = 0.12). The additive variance ($\:{\sigma\:}_{A}^{2}$) was larger than the dominance variance ($\:{\sigma\:}_{D}^{2}$) for NPP, PDW, SDT, and HSW. In contrast, the dominance component accounted for the majority of genetic variance for other traits. A low genotypic coefficient of variation (CV_g) was observed for most traits. The highest CV_g value (45.45%) was found for EMR, followed by YPP (36.34%) and GY (34.15%). Higher phenotypic coefficients of variation (CV_p) were recorded for the same traits and NPP. A small difference was found between CV_g and CV_p for phenological traits DTF and GFD, seed traits SDL, SDW, SDT and HSW and PDW, showing that the environment effect on these traits was less compared to other traits. On the other hand, the largest CV_g–CV_p difference was observed for NPP, followed by YPP and GY, indicating a higher environmental effect on these traits. Broad-sense heritability (H²) estimates ranged from 0 (NPP) to 0.76 (EMR). Most traits were moderately (NBR, PDL, PDW, PHE, GFD, SDT, YPP, and GY) to highly (EMR, DTF, SDL, SDW, HSW) heritable except NPP, NSP, GFD, SDT, and PDW which showed low broad-sense heritability (H² < 0.30). Narrow-sense heritability (h²) estimates were null for NPP, PHE, and NSP. In addition, low h² values were found for YPP, GY, GFD, and pod traits such as PWD and PDL. In contrast, DTF, EMR, and seed traits (SDL, SDW, SDT, and HSW) had moderate h² values ranging from 0.33 to 0.58 (Table 3). Estimates of expected genetic gain ranged from low (0.32%) for NPP to high (84.09%) for EMR. The magnitude of genetic gain was also high for YPP (48%) and GY (44.19%) while moderate expected gain was found for NBR (31.64%), PDL (17.58%), and PHE (18.18%). Low genetic gain (< 15%) was recorded for all other traits (Table 3).

Table 3

Experimental accuracy, variance components and genetic parameters estimates of 15 kersting’s groundnut yield-related traits
Traits	𝜎²_A	𝜎²_D	𝜎²_b(r)	𝜎²_𝜀	r_gg	CV_g	CV_p	h²	H²	GA	EGG
EMR	0.014	0.028	0.002	0.012	0.96	45.45	50.82	0.51	0.76	0.37	84.09
NBR	1.129	2.915	0.002	3.724	0.87	21.24	29.46	0.23	0.52	2.99	31.64
DTF	1.274	2.383	0.591	1.292	0.95	3.91	4.82	0.40	0.66	3.20	6.55
NPP	6.204	0.004	596.649	1296.652	0.12	2.73	47.78	0.00	0.00	0.29	0.32
PDL	0.274	1.837	0.254	1.602	0.89	11.69	16.03	0.13	0.53	2.18	17.58
PDW	0.033	0.021	0.023	0.064	0.85	2.87	4.62	0.27	0.38	0.29	3.72
PHE	0.000	0.014	0.003	0.022	0.79	14.23	23.59	0.00	0.36	0.14	18.18
NSP	0.000	0.006	0.007	0.023	0.71	7.75	17.33	0.00	0.21	0.09	6.98
GFD	0.550	0.790	0.060	2.031	0.82	2.02	3.23	0.21	0.39	1.49	2.60
SDL	0.074	0.072	0.003	0.052	0.95	4.56	5.32	0.58	0.73	0.66	8.05
SDW	0.010	0.030	0.000	0.020	0.93	3.52	4.31	0.33	0.67	0.34	5.99
SDT	0.021	0.002	0.011	0.031	0.83	3.48	5.85	0.33	0.35	0.18	4.23
HSW	0.961	0.302	0.154	0.562	0.93	8.71	10.89	0.57	0.64	1.85	14.35
YPP	0.149	17.258	4.707	20.404	0.85	36.35	56.8	0.01	0.41	5.51	48.00
GY	778.830	29133.602	7329.883	38794.607	0.84	34.15	54.45	0.02	0.39	223.79	44.19
𝜎²_A: additive variance, 𝜎²_D: dominance variance, 𝜎²_b(r): variance of block nested in replication, 𝜎²_𝜀: residual variance, r_gg: experimental accuracy, CV_g: genotypic coefficient of variation, CV_p: phenotypic coefficient of variation, h²: narrow-sense heritability, H²: broad-sense heritability, GA: genetic advance, EGG: excepted genetic gain. EMR: emergence rate, NBR: number of branches, DTF: days to 50% flowering, GFD: grain filling duration, NPP: number of pods per plant, PDL: pod length, PDW: pod width, PHE: pod harvest efficiency, NSP: number of seeds per pod, SDT: seed thickness, SDW: seed width, SDL: seed length, HSW: 100-seed weight, YPP: yield per plant, GY: grain yield.

[Please insert Table 3 here]

Structural equation modeling of yield components

A structural equation model (SEM) was developed to clarify the relationships between grain yield and yield components in kersting’s groundnut. The initial model did not adequately fit the dataset (RMSEA = 0.086, CFI = 0.822, and 𝜒² p-value = 0.021). Modification indices suggested a direct path from number of branches (NBR) to pod harvest efficiency (PHE), resulting in an adequate fit (RMSEA = 0.026, CFI = 0.999, and 𝜒² p-value = 0.325). Significant paths and variance explained (R²) by the final model for the four endogenous variables are shown in Fig. 1. The model explained 97% of the variation in grain yield (GY) while only 36% of the variation of yield per plant (YPP) was explained by the postulated relationships (Fig. 1). The direct and indirect effects of each yield component on YPP and GY are shown in Table 4. Number of pods per plant (NPP) had a substantial large positive effect on YPP and GY, with a high direct effect (0.43) on YPP and a large indirect effect (0.47) on GY (Fig. 1; Table 4). Number of seeds per pod (NSP) was the second more positively contributing factor to YPP and GY, with high direct (0.38) and indirect (0.40) effects on YPP and GY, respectively. Hundred seed weight (HSW) also had a positive and moderate effect on both traits. The effect of emergence rate (EMR) and seed thickness (SDT) was moderate and negative on both YPP and GY. The linear relationships between number of branches (NBR), days to 50% flowering (DTF), grain filling duration (GFD), pod width (PDW), seed length (SDL), and YPP were non-significant in the SEM model (Table 5). A similar trend was observed with GY, except NBR which had a small but significant direct effect (0.03), and pod length (PDL) as well as seed width (SDW) which were non-significant (Table 4). It is worth noting that NBR also showed a positive significant (P < 0.05) direct effect on HSW (Fig. 1).

Table 4

Direct, indirect and total effects of yield components on yield per plant and grain yield in kersting’s groundnut
Yield components	Yield per plant (YPP)			Grain yield (GY)
Yield components	Direct effects	Indirect effects	Total effects	Direct effects	Indirect effects	Total effects
EMR	-0.252***	0.000	-0.252	0.021	-0.274***	-0.253
NBR	-0.019	-0.026	-0.045	0.034**	-0.022	0.012
DTF	-0.028	-0.017	-0.045	-0.005	-0.031	-0.037
GFD	0.054	0.020	0.074	-0.023	0.060	0.036
NPP	0.427***	0.078**	0.505	0.005	0.466***	0.471
NSP	0.377**	-0.109*	0.269	0.016	0.404***	0.421
PDL	0.091	0.114*	0.205	-0.046	0.105	0.059
PDW	-0.304	-0.020	-0.324	0.026	-0.329	-0.303
PHE	0.145*	0.000	0.145	0.000	0.157*	0.158
SDL	0.017	0.051	0.068	0.019	0.022	0.041
SDW	0.117	0.059*	0.176	0.012	0.132	0.143
SDT	-0.191**	0.029	-0.162	0.007	-0.206***	-0.199
HSW	0.229*	0.000	0.229	0.016	0.249*	0.265
Significance level: p ≤ 0.05, p ≤ 0.01, **p ≤ 0.05. EMR: emergence rate, NBR: number of branches, DTF: days to 50% flowering, GFD: grain filling duration, NPP: number of pods per plant, PDL: pod length, PDW: pod width, PHE: pod harvest efficiency, NSP: number of seeds per pod, SDT: seed thickness, SDW: seed width, SDL: seed length, HSW: 100-seed weight, YPP: yield per plant, GY: grain yield.

Table 5

Selection differentials for Kersting’s groundnut yield component traits
Traits	Factor	Goal	Xo	Xs	SD (%)
EMR	FA5	Increase	0.44 ± 0.01	0.51	11.67
NBR	FA1	Increase	9.45 ± 0.10	9.65	1.09
DTF	FA3	Decrease	48.82 ± 0.10	48.56	-0.35
NPP †	-	Increase	91.21 ± 2.64	-	-
PDL	FA4	Increase	12.40 ± 0.06	12.93	2.28
PDW	FA4	Increase	7.80 ± 0.01	7.85	0.27
PHE	FA2	Increase	0.77 ± 0.00	0.78	0.55
NSP	FA4	Increase	1.29 ± 0.00	1.31	0.32
GFD	FA3	Decrease	57.26 ± 0.07	57.08	-0.12
SDL	FA1	Increase	8.20 ± 0.02	8.29	0.73
SDW	FA1	Increase	5.68 ± 0.01	5.75	0.81
SDT	FA5	Increase	4.26 ± 0.00	4.29	0.21
HSW	FA1	Increase	12.89 ± 0.07	13.18	1.42
YPP	FA2	Increase	11.48 ± 0.14	11.87	1.24
GY	FA2	Increase	506.40 ± 5.61	519.91	0.90
† Trait not included in the selection index due to null heritability and constant best linear unbiased prediction (BLUP) values. Xo: Genetic value, Xs: selected value, SD: selection differential, FA: factor analysis, EMR : emergence rate, NBR : number of branches, DTF : days to 50% flowering, GFD : grain filling duration, NPP : number of pods per plant, PDL : pod length, PDW : pod width, PHE : pod harvest efficiency, NSP : number of seeds per pod, SDT : seed thickness, SDW : seed width, SDL : seed length, HSW : 100-seed weight, YPP: yield per plant, GY: grain yield.

[Please insert Fig. 1 here]

[Please insert Table 4 here]

Multi-trait selection

Trait associations and trait profiles in kersting’s groundnut accessions

The genotype by trait (GT) biplot represents 50.51% of the variation of the 81 accessions for the 15 traits (Fig. 2a). Grain yield (GY) was negatively correlated with SDT, DTF, and NBR indicating that high-yielding accessions were characterized by thin seeds, early flowering, and low number of branches. On the other hand, GY was positively correlated with YPP, NPP, PHE, NSP, SDL, SDW, HSW, PDW, and GFD. There was no correlation between GY and EMR. The trait vector angles showed that the strength of correlation was extremely high with YPP, and high with NPP and PHE. Moderate (NSP, SDL, SDW, HSW, SDT, DTF, and NBR) to low (PWD and GFD) correlations existed between GY and the other traits. Figure 2a also shows the trait profiles of the accessions. Accessions such as BUR8, BUR18, BUR14, BUR16, ZKU, and ZHLA2 had high yield, number of pods per plant, and pod harvest efficiency. Moreover, they had larger, longer, and heavier seeds but their seeds were also flat, and they flowered early.

The genotype by yield*trait (GYT) biplot accounted for 94.44% of the total variation in the dataset (Fig. 2b). All yield-trait combinations appeared to be positively correlated with one another as indicated by the acute angles (< 90°) between their vectors. Strong correlations were found between GY*YPP, GY*SDW, GY*SDL, and GY*NSP, indicating that the concurrent selection for YPP, SDW, SDL, and NSP can help increase yield in some accessions (ZKU, GBO2). Similarly, there were high correlations between GY*PHE, GY*SDT, GY*PDW, and GY*NPP, suggesting high suitability of a combination of PHE, SDT, PDW, and NPP with GY for improving the productivity of genotypes such as AGN1 (Fig. 2b). Moreover, the GYT biplot shows that the combination of HSW and PDL with GY will favor the enhancement of productivity in some well-performing genotypes (ZHLA2, BUR14).

Figure 2c is the polygon or “which-won-where” view of the GYT biplot. The yield-trait combinations were divided into two sectors; the first one corresponding to GY*NBR while the second sector was associated with the remaining yield-trait combinations. The genotype BUR18 had the largest values for most yield-trait combinations except GY*NBR. It was followed by GBO2, and to some extent by BUR8, BUR16, and BUR7 which had comparable performances. In contrast, HAY2 followed by ALI1 had the highest values for GY*NBR suggesting that these genotypes were best in combining GY with NBR (Fig. 2c).

Figure 2d ranked genotypes based on their overall superiority as well as their strengths and weaknesses. The best accessions were by order of superiority BUR18 > BUR8 > BUR14 > GBO2 > ZHLA2 > ZHU. The accessions ZKE, KNO1, and VIV were ranked the poorest overall. Moreover, the figure shows that BUR18 and BUR8 exhibited high values for traits such as pod length (PDL), seed size (SDW and SDL) and weight (HSW and YPP), NSP as well as early flowering (DTF). GBO2 exhibited higher values for pod productivity (NPP) and harvest efficiency (PHE), PDW, SDT, and early maturity (GFD). The accessions BUR14, ZHLA2, and ZKU displayed well-balanced performance across the various traits.

[Please insert Fig. 2 here]

Multi-trait genotype-ideotype distance index

The MGIDI index was used to select superior genotypes based on all 15 target traits. Overall, the index provided a desirable selection differential (SD) for all traits except NPP which was not included in the index due to its null heritability value and constant BLUP values. The SD values obtained from the MGIDI index were generally low across all traits (Table 5). For traits in which high values were desired, SD ranged from 0.21% (SDT) to 11.67% (EMR), indicating the potential for selecting genotypes with favorable characteristics. Conversely, for traits in which lower values were preferred, SD varied from − 0.12% (GFD) to -0.35% (DTF), indicating the possibility of selecting genotypes with reduced values in these traits. The total genetic gains were higher for traits where an increase was desired, with gains up to 21.49%. Meanwhile, gains were lower for traits where a decrease was preferred, with total gains as low as -0.47%.

Figure 3a shows the genotype ranking according to the MGIDI index. Based on the 15% selection pressure, 12 genotypes were selected whose performance was closer to that of the ideotype. The selected genotypes included ENA2, DOG, ZHLA1, ITK2, BUR8, BUR3, ZHLA3, BUR7, BIN, ODM2, GBO5 and BOD (Fig. 3a). Conversely, the genotypes BUR12 and KAH2 performed the poorest in comparison to the ideotype. Figure 3b presents the strengths and weaknesses of the selected genotypes. In this figure, the factors that contributed the most to genotype selection via the MGIDI index were placed close to the plot center, while traits contributing the least were positioned near the periphery. The selected genotypes exhibited strengths in traits such as pod harvesting efficiency (PHE), yield per plant (YPP), and grain yield (GY) (Fig. 3b; Table 5). Moreover, eight out of the 12 genotypes exhibited strengths towards pod length (PDL), pod width (PDW), and number of seeds per pod (NSP; Fig. 3b). Seed thickness and emergence rate were found more contributing towards the selection of some genotypes (BUR8, BIN).

[Please insert Table 5 here]

[Please insert Fig. 3 here]

The estimation of genetic parameters of target traits is crucial for successful breeding, especially for quantitative traits with complex genetics such as yield. Dissecting the genetic variance can provide valuable insights into the transmissibility of these traits and inform breeders on the most effective breeding strategies. However, this process can be challenging and requires a significant amount of resources and time ¹⁷, especially for orphan crops ^66–68. This study employed an approach based on SNP markers to effectively partition the genetic variance and estimate the narrow sense heritability of 15 quantitative traits in kersting’s groundnut.

The results revealed moderate broad sense heritability for yield per plant (YPP) and grain yield (GY), with H² values of 0.41 and 0.39, respectively, which is consistent with a previous study (H² = 0.43 for GY) ¹⁵. However, the narrow sense heritability (h²) estimates for YPP and GY were near zero, so were the estimates for other traits like number of pods per plant (NPP), pod harvest efficiency (PHE), and number of seeds per pod (NSP). This suggests that the observed variation among accessions was largely due to environmental variation and dominance effects. Thus, improving kersting’s groundnut yield through conventional breeding methods may be challenging due to the low additive genetic variance. Despite the low h² estimates, the high genetic advance (> 30%) for YPP and GY suggests that these traits are to some extent under genetic control and can be improved through selection, although progress may be slow. However, moderate H² and high genetic gain for yield are not sufficient conditions to ensure the successful breeding of a crop ¹⁰. Other factors such as genetic diversity, reproductive biology, as well as genetics and environmental factors influencing the target trait must be considered. To increase the chances of success in identifying superior kersting’s groundnut genotypes, the working germplasm should be expanded to perhaps a couple of thousands of accessions, which requires developing high-throughput phenotyping methods to evaluate a large number of genotypes in a short period ^69,70.

Moderate narrow-sense (h² = 0.40 to 0.58) and high broad-sense (H² = 0.64 to 0.76) heritability estimates were obtained for traits such as 100-seed weight (HSW), seed length (SDL), days to 50% flowering (DTF) and emergence rate (EMR), indicating that these traits were highly heritable and there was enough genetic variation present in the population to allow for response to selection. Therefore, HSW, SDL, DTF, and EMR were likely to be effectively improved through direct or indirect selection using strongly correlated traits.

The result highlighting that NPP, NSP, and HSW affect YPP and GY in the same direction was expected and is consistent with the evidence in the legume breeding literature ^71–73. Interestingly, a substantial negative impact of NSP on HSW was observed. This observation implies the existence of an imbalanced trade-off in resource allocation between seed number and seed weight during the seed set period, where available resources are unequally distributed between the two yield components ⁷⁴. In the case of kersting’s groundnut, more seeds are seemingly competing for resources which reduces the resource each seed could be allocated, resulting in a negative effect on seed weight. Therefore, selecting genotypes with a high seed number per pod will result in low seed weight which would ultimately impact GY. Hence, it can be recommended that breeders perform a multi-trait selection for targeting GY, NPP, NSP, and HSW concurrently. The emergence rate (EMR) and seed thickness (SDT) also exhibited a significant negative indirect effect through YPP on which they both exerted a negative direct effect. This confirms the statement that grain yield is indirectly affected by a wide range of component traits, either negatively or positively through undetermined mechanisms ⁷⁵. Consequently, applying indirect selection for thicker kersting’s groundnut grains will result in a decrease in YPP which in turn will decrease GY. Although the effect of the number of branches (NBR) on GY was small (r = 0.03), it was significant, which implies there is, to some extent, a trade-off between reproductive and vegetative allocation in kersting’s groundnut. Therefore, NBR can be used for indirect selection for GY.

In many breeding programs, yield is considered the most important trait to determine the effectiveness of a genotype, while other traits including morphological, and physiological characteristics are only valuable when they are combined with high yield levels ^76,77. For instance, traits such as drought resistance, earliness, or disease resistance alone do not hold much significance to growers, suggesting that the selection of the best genotypes based on the combined effects of yield and yield-related traits will be more meaningful. The accessions BUR18 and BUR8 were best in combining grain yield with seed sizes and day to 50% flowering. Hence, they can be considered as suitable candidates for developing early flowering cultivars with high yield potential for target production areas with shorter growing seasons. Meanwhile, BUR14, ZHLA2, and ZKU displayed average performance across traits, and they will be suitable for most production areas, assuming they have a stable performance. Therefore, it is necessary to assess the stability of the selected accessions in the major growing environments to validate their superiority.

Overall, this study investigated the dissection of genetic variability for yield and yield-related traits, and multi-trait selection in kersting’s groundnut. Sufficient variations were observed among accessions for grain yield and all yield-related traits, except NSP and NPP. Most traits had moderate to high broad sense heritability, with seed-related traits (SDT, SDL, SDW, HSW), EMR, and DTF exhibiting moderate narrow sense heritability, making them suitable for recurrent selection. However, the low narrow sense heritability of YPP and GY indicates that direct selection is not efficient, but indirect selection based on significant component traits (NPP, NSP, HSW, SDT, EMR) would be effective. The genotype by yield*trait (GYT) biplot identified six accessions (BUR18, BUR8, BUR14, GBO2, ZHLA2 and ZHU) as superiors while the multi-trait genotype-ideotype distance (MGIDI) index selected 12 accessions (ENA2, DOG, ZHLA1, ITK2, BUR8, BUR3, ZHLA3, BUR7, BIN, ODM2, GBO5 and BOD) as promising parents for KG yield improvement.

Germplasm and experiment design

The germplasm used in this study consisted of 81 accessions from Benin (70) and Burkina Faso (11). Accessions from Benin were obtained directly from farmers with their informed consent, while those from Burkina Faso were provided by the genebank of the Institute of Environment and Agricultural Research of Burkina Faso (INERA), under a Material Transfer Agreement (MTA). All collections were conducted in compliance with the Nagoya Protocol on Access and Benefit-Sharing. Seeds from these accessions are stored in the Laboratory of Applied Ecology’s genebank in Benin. The experiment was carried out at the Djidja regional station of the National Institute of Agricultural Research of Benin (INRAB) (7°19'4.30" N, 1°55'0.65" E). Experiment was laid out in 9×9 alpha lattice design with three replications. Replicates consisted of 9 blocks, each containing 9 plots randomly assigned with the 81 accessions. Plots were 1 m apart and each plot contained three 4.5 m rows with 0.75 m between row spacing (10.125 m²/plot). Planting was done with 0.3 m within row spacing, giving 15 plants per row and 45 plants per plot for an approximate density of 44500 plants per hectare. The trial was rainfed, field was weeded three times, and no fertilizer was applied to the crop. Thalis 112 EC pesticide was applied at the rate of 225 ml/ha, when field was attacked by fungi.

Data collection

Fifteen quantitative agro-morphological traits (Table 6) were recorded on a plant or plot basis. Data collected include emergence rate (EMR), number of branches (NBR), days to 50% flowering (DTF), number of pods per plant (NPP), pod length (PDL), pod width (PDW), pod harvest efficiency (PHE), number of seeds per pod (NSP), grain filling duration (GFD), seed thickness (SDT), seed width (SDW), seed length (SDL), 100-seed weight (HSW), yield per plant (YPP) and grain yield (GY). Four traits viz. NBR, NPP, PHE and YPP were recorded on 10 randomly picked plants, while the remaining eleven traits were recorded on whole plot basis as described in Table 6 below.

Table 6

Quantitative traits recorded on the kersting’s groundnut accessions
Trait	Code	Data collection method
Emergence rate (%)	EMR	Determined the percentage of emerged seedling
Number of branches	NBR	Count of the number of branches per plant averaged from 10 random plants
Days to 50% flowering	DTF	Determined as the number of days from sowing to when 50% of plants in a plot had at least one flower
Number of pods per plant	NPP	Average count of the number of pods per plant from a sample of 10 plants
Pod length (mm)	PDL	Measured on 50 random pods
Pod width (mm)	PDW	Measured on 50 random pods
Pod harvest efficiency (%)	PHE	Determined as average percentage of pods with filled grain from a random sample of 10 plants
Number of seeds per pod	NSP	Average number of seeds per pod from a sample of 50 random pods
Grain filling duration (days)	GFD	Counted the number of days from 50% flowering to 90% of plant have mature pods
Seed length (mm)	SDL
Seed width (mm)	SDW	Measured as the average of three sets of 50 random seeds per plot
Seed thickness (mm)	SDT
100-seed weight (g)	HSW	Weight of 100 healthy seeds, calculated as the average of three sets of 100 seeds per plot.
Yield per plant (g plant^− 1)	YPP	Average weight of all seeds produced per plant from 10 random plants
Grain yield (Kg ha^− 1)	GY	Product of average grain yield per plant at ~ 12% moisture content and plant density

[Please insert Table 6 here]

Genotype data

Single nucleotide polymorphism (SNP) markers generated with DArTseq were used, SNP curation workflow is described in Kafoutchoni, et al. ⁵⁰. Briefly, missing markers were imputed using probabilistic principal component analysis (PPCA) as implemented in the Diversity Arrays Technology’s KD-Compute software (https://kdcompute.seqart.net/kdcompute). Moreover, SNPs with more than 60% missing data and accessions with more than 80% missing data were removed from the dataset. SNPs with read depth < 2, call rate < 75%, reproducibility < 95% and polymorphic information content (PIC) of zero were also removed. A set of 1117 SNP markers that passed quality tests were retained. The final genotypic data was converted into the dosage format (-1 = homozygotes for the reference-allele, 0 = heterozygotes, and 1 = homozygotes for the alternative-allele) and was used to construct additive (Ga) and dominance (Gd) relationship matrices.

Statistical analysis

Phenotypic analysis

All statistical analyses were performed in R 4.1.2 software ⁵¹. Descriptive statistics (mean, standard error, range, and coefficient of variation) were computed for each yield-related trait. Traits were subsequently subjected to a linear mixed model analysis in the sommer package ⁵² using the Direct-inversion Newton-Raphson (NR) restricted maximum likelihood (REML) algorithm ⁵³. The model structure is given by Eq. (1):

$$\:\varvec{y}=X\varvec{\beta\:}+V\varvec{r}+T\varvec{g}+\varvec{\epsilon\:}$$

where y is the vector of phenotypic observations; β is the vector of the fixed effects of replication added to the overall mean; $\:\varvec{r}$ is the effect of blocks nested within replications and was considered as random [$\:\varvec{r}\:\sim\:\:N\left(0,\:I{\sigma\:}_{b\left(r\right)}^{2}\right)$]; $\:\varvec{g}$ is the vector of accession effects and was regarded as random [$\:\varvec{g}\sim\:\:N\left(0,\:I{\sigma\:}_{g}^{2}\right)$]; $\:\varvec{\epsilon\:}$ is the vector for error [$\:\varvec{\epsilon\:}\sim\:\:N\left(0,\:I{\sigma\:}_{e}^{2}\right)$]; and $\:X$, $\:V$, and $\:T$ are the incidence matrices that relate the independent vectors to the response variable $\:\varvec{y}$. Wald’s and likelihood ratio tests were performed to assess the significance of fixed and random factors, respectively.

Based on the observation that when molecular markers are available, no specific mating design is required to dissect the genetic variance into its additive and non-additive (i.e., dominance and epistatic) components ⁵², the model in Eq. (1) was extended to the following variant (Eq. (2), which enabled to capture the additive ($\:{\sigma\:}_{A}^{2}$) and dominance ($\:{\sigma\:}_{D}^{2}$) genetic variances:

$$\:\varvec{y}=X\varvec{\beta\:}+\text{V}\varvec{r}+{Z}_{add}\varvec{a}+{Z}_{dom}\varvec{d}+\varvec{\epsilon\:}$$

where $\:\varvec{y}$ is raw phenotypic observations; $\:\varvec{\beta\:}$ is the vector of the fixed effects of replication; $\:\varvec{r}$ is the random effect of blocks nested within replications [$\:\varvec{r}\:\sim\:\:N\left(0,\:I{\sigma\:}_{b\left(r\right)}^{2}\right)$] ; $\:\varvec{a}$ is the additive effect [$\:\varvec{a}\sim\:\:N\left(0,\:{G}_{a}{\sigma\:}_{A}^{2}\right)$], where $\:{G}_{a}$ is the additive genetic relationship matrix; $\:\varvec{d}$ is the dominance effect [$\:\varvec{d}\sim\:\:N\left(0,\:{G}_{d}{\sigma\:}_{D}^{2}\right)$], where $\:{G}_{d}$ is the dominance genetic relationship matrix; $\:\varvec{\epsilon\:}$ is the vector for error [$\:\varvec{\epsilon\:}\sim\:\:N\left(0,\:I{\sigma\:}_{e}^{2}\right)$]; and $\:V$, $\:{Z}_{add}$, and $\:{Z}_{dom}$ are the incidence matrices relating observations to the levels of each factor. The $\:{G}_{a}$ and $\:{G}_{d}$ matrices were obtained from SNP data respectively using the functions A.mat and D.mat available in the sommer package.

Broad-sense (H²) and narrow-sense (h²) heritabilities were estimated from the aforementioned model to capture the fraction of total phenotypic variance due to genotypic and additive variation, respectively. H² and h² were estimated respectively using the following equations (3) and (4) ⁵⁴:

$$\:{H}^{2}=\frac{{\sigma\:}_{A}^{2}+{\sigma\:}_{D}^{2}}{{\sigma\:}_{A}^{2}+{\sigma\:}_{D}^{2}+{\sigma\:}_{b\left(r\right)}^{2}+{\sigma\:}_{e}^{2}}$$

$$\:{h}^{2}=\frac{{\sigma\:}_{A}^{2}}{{\sigma\:}_{A}^{2}+{\sigma\:}_{D}^{2}+{\sigma\:}_{b\left(r\right)}^{2}+{\sigma\:}_{e}^{2}}$$

where $\:{\sigma\:}_{A}^{2}$ is the additive variance component; $\:{\sigma\:}_{D}^{2}$ is the dominance variance component; $\:{\sigma\:}_{b\left(r\right)}^{2}$ is the variance component related to the effect of incomplete blocks nested within replications; and $\:{\sigma\:}_{e}^{2}$ is the residual variance component.

The expected genetic gain was also estimated from the same model in Eq. (2) as the percentage of genetic advance over the population mean following Johnson, et al. ⁵⁵ formula (Eq. (5):

$$\:EGG=\:\frac{GA}{\mu\:}\times\:100$$

Where GA is the genetic advance expressed as $\:GA=\:\frac{K{\sigma\:}_{g}^{2}}{\sqrt{{\sigma\:}_{p}^{2}}}$, with $\:{\sigma\:}_{g}^{2}$ = genotypic variance, $\:{\sigma\:}_{p}^{2}$ = phenotypic variance and K = Selection differential at 5% selection pressure i.e., 2.063. µ is the population mean.

Finally, to assess data quality, the experimental accuracy was computed using the Eq. (6) below:

$$\:{r}_{gg}={(1-\frac{1}{1+{rCV}_{R}^{2}})}^{1/2}$$

Where r is the number of replicates, $\:{CV}_{R}={CV}_{g}/{CV}_{e}$_, with $\:{CV}_{g}=\left(\frac{\sqrt{{\sigma\:}_{g}^{2}}}{\mu\:}\right)\bullet\:100$ and $\:{CV}_{e}=\left(\frac{\sqrt{{\sigma\:}_{e}^{2}}}{\mu\:}\right)\bullet\:100$, and µ is the grand mean of the trait ⁵⁶.

Structural equation modeling

A structural equation model (SEM) was performed in the lavan R package ⁵⁷, to quantify the multivariate causal relationships network among grain yield and fourteen potential yield components. SEM requires to specify an initial model that hypothesizes the causal relationships among the variables at hand ⁵⁸. Therefore, an initial model was specified including relationships confirmed in previous studies on legume crops such as chickpea ⁵⁹, cowpea ⁶⁰ and kersting’s groundnut ¹⁵. Four yield component traits viz. PHE, HSW, YPP, and GY were considered as endogenous variables while the remaining 11 traits were considered as exogenous variables in the initial model. The model was subsequently modified using the modification indices until an acceptable model fit was achieved. Modification indices were a standard metric for improving SEM model fit ³². The comparative fit index (CFI), root mean square error of approximation (RMSEA), and Chi-square p-value were used to assess the model goodness of fit. Models with a CFI > 0.95, RMSEA < 0.05, and p-value > 0.05 were considered to fit well. Standardized partial regression coefficients were reported to permit direct comparisons across paths ⁶¹. Standardized coefficients with absolute values below 0.10 are often considered to have a small effect, while values around 0.30 show a medium effect and values above 0.50 indicate a large effect ⁶¹.

Multi-trait genotype selection

To reveal the associations between traits and trait profiles (i.e., strengths and weaknesses) of accessions, genotype by trait (GT) ⁶², and genotype by Yield*Trait (GYT) ⁶³, biplots were generated. For this purpose, a two-way genotype by trait matrix consisting of the 81 accessions (in rows) and the 15 traits (in columns) was constructed by extracting the best linear unbiased predictions (BLUPs) of all traits from the previous mixed-effect models. BLUPs were extracted using the emmeans R package ⁶⁴. The GYT table was subsequently obtained by multiplying the grain yield (GY) value with each of the other trait values, for each accession. For traits such as days to 50% flowering (DTF) and grain filling duration (GFD) for which lower values are desirable, the values for the yield-trait combinations were obtained by dividing the GY value with the trait value ⁶³. The datasets were then standardized to a mean of 0 and a unit variance and the GT and GYT biplots were constructed in the metan package ⁶⁵ based on the first two principal components.

The multi-trait genotype-ideotype distance index (MGIDI) was used to rank the accessions based on desired values of multiple traits as proposed by Olivoto and Nardino ⁴⁵. The steps to compute the MGIDI index were fourfold.

In the first step, the traits were rescaled to a 0–100 range such a way that 0 and 100 represent the minimum and maximum of the traits in which positive gains are desired, and the inverse for traits in which negative gains are desired.
In the second step, an exploratory factor analysis was performed to group correlated traits into factors and estimate factorial scores for each genotype.
The third step was ideotype planning, where the ideotype had the highest rescaled value (100) for all considered traits and was defined by a [1×p] vector.
In the fourth step, the Euclidean distance between the scores of the genotypes and the ideotype was computed as the MGIDI index using the following Eq. (7) ⁴⁵:

$$\:{MGIDI}_{i}={\left[\sum\:_{j=1}^{f}{({\gamma\:}_{ij}-{\gamma\:}_{j})}^{2}\right]}^{0.5}$$

Where γ_ij is the score of the ith genotype in the jth factor (i = 1, 2, ..., g; j = 1, 2, ..., f), where g and f are the number of genotypes and factors, respectively; and γ_j is the jth score of the ideotype.

The genotype with the lowest MGIDI is closer to the ideotype representing desired values for all the studied traits. The selection differential was calculated for all traits considering a selection intensity of 15%. Data manipulation and computation of the MGIDI index were performed in the metan R package ⁶⁵.

Competing interests Statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

This work was supported by the Netherlands Organisation for Scientific Research (NWO-WOTRO) [grant number W08.270.344]; the Regional Universities Forum for Capacity Building in Agriculture (RUFORUM) [grant number RU/2018/TQA/38]; the World Academy of Sciences [Grant number 18–238 RG/BIO/AF/AC_G-FR3240303667]; and the AGNES-BAYER Science Foundation and the Alexander von Humboldt Foundation (AvH).

Author Contribution

KMK: Conceptualization, Data curation, Formal analysis, Funding acquisition, Writing – original draft, Writing – review & editing; EEA: Conceptualization, Funding acquisition, Project administration, Resources, Writing – review & editing; HSS: Data curation, Writing – review & editing; GG: Formal analysis, Writing – review & editing; AEA: Funding acquisition, Resources, Supervision; CA: Resources, Supervision; FAKS: Writing – review & editing; SA: Writing – review & editing.

Acknowledgement

We thank the farmers who provided the accessions used in this study. Special thanks to Mr. Sergino Ayi (B.Sc.) and Mr. Yelognissè Gilles Chodaton (M.Sc.) for their assistance during the field work.

Data Availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Dansi, A. et al. Diversity of the neglected and underutilized crop species of importance in Benin. The scientific world journal 2012, doi:10.1100/2012/932947 (2012).
Chikwendu, J. N. Comparative evaluation of chemical composition of fermented ground bean flour (Kerstingella geocarpa), cowpea flour (Vigna unguiculata) and commercial wheat flour (Triticum spp.). Pakistan Journal of Nutrition 14, 218–224 (2015).
Echendu, A., Obizoba, I., Ngwu, E. & Anyika, J. Chemical composition of ground bean based cocoyam, yam and plantain pottage dishes and roasted ground bean. Pakistan Journal of Nutrition 8, 1786–1790 (2009).
Fassinou, F. T. K. et al. Comparative Analysis of Proximate Compositions, Physical Properties, and Sensory Attributes of Kersting’s Groundnut (Macrotyloma geocarpum (Harms) Maréchal & Baudet) Accessions. J. Food Qual. 2023, 6372248, doi:10.1155/2023/6372248 (2023).
Coulibaly, M. et al. Profiling Cultivars Development in Kersting's Groundnut [Macrotyloma geocarpum (Harms) Maréchal and Baudet] for Improved Yield, Higher Nutrient Content, and Adaptation to Current and Future Climates. Frontiers in Sustainable Food Systems 5, doi:10.3389/fsufs.2021.759575 (2022).
Agoyi, E. E. et al. Kersting’s Groundnut [Macrotyloma geocarpum (Harms) Maréchal & Baudet] crop attracts more field pests and diseases than reported before. Agricultural Research & Technology Open Access Journal 21 (2019).
Agoyi, E. E., Sossou, H. S., Sodédji, F. A. K., Assogbadjo, A. E. & Sinsin, B. DOYIWÉ Kersting’s Groundnut [Macrotyloma Geocarpum (Harms) Maréchal & Baudet] A Versatile & Underutilized Grain Legume for Subhumid Zones. Recommended Practices for Farmers. (2020).
Agoyi, E. E., Kafoutchoni, K. M. & Ayena, M. A. T. in Neglected and Underutilized Crops (eds Muhammad Farooq & Kadambot H. M. Siddique) Ch. 14, (Academic Press, 2023).
Majidi, M. M., Hoseini, B., Abtahi, M., Mirlohi, A. & Araghi, B. Genetic analysis of seed related traits in Orchardgrass (Dactylis glomerata) under normal and drought stress conditions. Euphytica 203, 409–420 (2015).
Acquaah, G. Introduction to quantitative genetics. Principles of Plant Genetics and Breeding, Second Edition, 63–94 (2012).
Ogunniyan, D. J. & Olakojo, S. A. Genetic variation, heritability, genetic advance and agronomic character association of yellow elite inbred lines of maize (Zea mays L.). Nigerian Journal of Genetics 28, 24–28, doi:https://doi.org/10.1016/j.nigjg.2015.06.005 (2014).
Hamidou, M. et al. Genetic variability and its implications on early generation sorghum lines selection for yield, yield contributing traits, and resistance to sorghum midge. International Journal of Agronomy 2018 (2018).
Onwubiko, N. C., Uguru, M. I. & Chimdi, G. O. Estimates of genetic parameters in bambara groundnut {Vigna subterranea (L.) VERDC.}. Plant Breeding and Biotechnology 7, 295–301 (2019).
Peñuela, M. et al. Is the variability in my crop due to genetic or environmental factors? Using heritability to solve this question. Journal of Crop Science and Biotechnology 25, 161–172 (2022).
Akohoue, F., Achigan-Dako, E. G., Coulibaly, M. & Sibiya, J. Correlations, path coefficient analysis and phenotypic diversity of a West African germplasm of Kersting’s groundnut [Macrotyloma geocarpum (Harms) Maréchal & Baudet]. Genetic Resources and Crop Evolution 66, 1825–1842 (2019).
Yang, J. et al. Common SNPs explain a large proportion of the heritability for human height. Nature genetics 42, 565–569 (2010).
Wolfe, M. D., Kulakow, P., Rabbi, I. Y. & Jannink, J.-L. Marker-based estimates reveal significant nonadditive effects in clonally propagated cassava (Manihot esculenta): implications for the prediction of total genetic value and the selection of varieties. G3: Genes, Genomes, Genetics 6, 3497–3506 (2016).
Weissbrod, O., Flint, J. & Rosset, S. Estimating SNP-based heritability and genetic correlation in case-control studies directly and with summary statistics. The American Journal of Human Genetics 103, 89–99 (2018).
Vogler, C. et al. Substantial SNP-based heritability estimates for working memory performance. Translational psychiatry 4, e438-e438 (2014).
Baselmans, B. M., Yengo, L., van Rheenen, W. & Wray, N. R. Risk in relatives, heritability, SNP-based heritability, and genetic correlations in psychiatric disorders: a review. Biological Psychiatry 89, 11–19 (2021).
Kostem, E. & Eskin, E. Improving the Accuracy and Efficiency of Partitioning Heritability into the Contributions of Genomic Regions. The American Journal of Human Genetics 92, 558–564, doi:https://doi.org/10.1016/j.ajhg.2013.03.010 (2013).
Pour-Aboughadareh, A. et al. Identification of salt-tolerant barley genotypes using multiple-traits index and yield performance at the early growth and maturity stages. Bulletin of the National Research Centre 45, 1–16 (2021).
Oladosu, Y. et al. Genotypic and phenotypic relationship among yield components in rice under tropical conditions. BioMed research international 2018 (2018).
Roy, R. K., Majumder, R. R., Sultana, S., Hoque, M. & Ali, M. Genetic variability, correlation and path coefficient analysis for yield and yield components in transplant aman rice (Oryza sativa L.). Bangladesh Journal of Botany 44, 529–535 (2015).
Shrestha, N. et al. Correlation Coefficient and Path Analysis of Advance Rice Genotypes in Central Mid-hills of Nepal. International Journal of Research in Agricultural Sciences 5, 2348–3997 (2018).
Alsabah, R., Purwoko, B., Dewi, I. & Wahyu, Y. Selection index for selecting promising doubled haploid lines of black rice. SABRAO Journal of Breeding and Genetics 51, 430–441 (2019).
Islam, M. R. et al. Selection index for genetic improvement of wheat (Triticum aestivum L.). Journal of Chemical, Biological and Physical Sciences (JCBPS) 7, 1 (2016).
Ibrahim, E. A., Abdalla, A. W. H., Rahman, M. A. & El Naim, A. M. Path coefficient and selection indices in sixteen guar (Cyamopsis Tetragonoloba L.) genotypes under rainfed. Int. J. Agric. For 2, 79–83 (2012).
Solomon, G. Correlation and path analysis in yield and yield components in spring bread wheat (Triticum aestivum L.) genotypes under irrigated condition in Southern India. African Journal of Agricultural Research 8, 3186–3192 (2013).
Kishore, N. S., Srinivas, T., Nagabhushanam, U., Pallavi, M. & Sameera, S. Genetic variability, correlation and path analysis for yield and yield components in promising rice (Oryza sativa L.) genotypes. SAARC Journal of Agriculture 13, 99–108 (2015).
Mason, S., Kmail, Z., Galusha, T. & Jukić, Ž. Path analysis of drought tolerant maize hybrid yield and yield components across planting dates. Journal of Central European Agriculture 20, 194–207 (2019).
Lamb, E., Shirtliffe, S. & May, W. Structural equation modeling in the plant sciences: An example using yield components in oat. Canadian Journal of Plant Science 91, 603–619, doi:10.4141/cjps2010-035 (2011).
Kent, M. A., Crozier, D. S. & Rooney, W. L. Assessment of kernel characteristics to predict popping performance in grain sorghum. Crop Science 62, 1051–1059 (2022).
Petraitis, P. S., Dunham, A. E. & Niewiarowski, P. H. Inferring Multiple Causality: The Limitations of Path Analysis. Functional Ecology 10, 421–431, doi:10.2307/2389934 (1996).
Mańkowski, D. R., Kozdój, J. & Janaszek-Mańkowska, M. Structural equation model as a tool to assess the relationship between grain yield per plant and yield components in doubled haploid spring barley lines (Hordeum vulgare L.). Plant Breeding and Seed Science 73, 63–77 (2016).
Du, Y.-L. et al. Yield components, reproductive allometry and the tradeoff between grain yield and yield stability in dryland spring wheat. Field Crops Research 257, 107930 (2020).
Heineck, G. C. et al. Relationships and influence of yield components on spaced-plant and sward seed yield in perennial ryegrass. Grass Forage Sci. 75, 424–437 (2020).
Santana, D. C. et al. Structural equation modelling and factor analysis of the relationship between agronomic traits and vegetation indices in corn. Euphytica 218, 1–8 (2022).
Céron-Rojas, J. J. & Crossa, J. Linear selection indices in modern plant breeding. (Springer Nature, 2018).
Smith, H. F. A discriminant function for plant selection. Annals of eugenics 7, 240–250 (1936).
Hazel, L. N. The genetic basis for constructing selection indexes. Genetics 28, 476–490 (1943).
Pesek, J. & Baker, R. Comparison of tandem and index selection in the modified pedigree method of breeding self-pollinated species. Canadian Journal of Plant Science 49, 773–781 (1969).
Williams, J. The evaluation of a selection index. Biometrics 18, 375–393 (1962).
Bizari, E. H., Val, B. H. P., Pereira, E. d. M., Mauro, A. O. D. & Unêda-Trevisoli, S. H. Selection indices for agronomic traits in segregating populations of soybean. Revista Ciência Agronômica 48, 110–117 (2017).
Olivoto, T. & Nardino, M. MGIDI: toward an effective multivariate selection in biological experiments. Bioinformatics 37, 1383–1389 (2021).
Pour-Aboughadareh, A. & Poczai, P. Dataset on the use of MGIDI index in screening drought-tolerant wild wheat accessions at the early growth stage. Data in Brief 36, 107096 (2021).
Nardino, M. et al. Understanding drought response mechanisms in wheat and multi-trait selection. PloS one 17, e0266368 (2022).
Uddin, M. S. et al. High-Throughput Root Network System Analysis for Low Phosphorus Tolerance in Maize at Seedling Stage. Agronomy 11, 2230 (2021).
Olivoto, T. et al. Multi-trait selection for mean performance and stability in maize. Agronomy Journal 113, 3968–3974 (2021).
Kafoutchoni, K. M., Agoyi, E. E., Agbahoungba, S., Assogbadjo, A. E. & Agbangla, C. Genetic diversity and population structure in a regional collection of Kersting’s groundnut (Macrotyloma geocarpum (Harms) Maréchal & Baudet). Genetic Resources and Crop Evolution 68, 3285–3300, doi:https://doi.org/10.1007/s10722-021-01187-4 (2021).
R: A language and environment for statistical computing (R Foundation for Statistical Computing, Vienna, Austria, 2022).
Covarrubias-Pazaran, G. Genome-assisted prediction of quantitative traits using the R package sommer. PloS one 11, e0156744 (2016).
Lee, S. H. & Van der Werf, J. H. MTG2: an efficient algorithm for multivariate linear mixed model analysis based on genomic information. Bioinformatics 32, 1420–1422 (2016).
Piepho, H.-P. & Möhring, J. Computing heritability and selection response from unbalanced plant breeding trials. Genetics 177, 1881–1888 (2007).
Johnson, H. W., Robinson, H. F. & Comstock, R. E. Estimates of genetic and environmental variability in soybeans. Agronomy journal 47, 314–318 (1955).
de Resende, M. D. V. & Duarte, J. B. Precision and quality control in variety trials. Pesquisa Agropecuária Tropical 37, 182 (2007).
Rosseel, Y. lavaan: An R package for structural equation modeling. Journal of statistical software 48, 1–36 (2012).
Grace, J. B. & Keeley, J. E. A structural equation model analysis of postfire plant diversity in California shrublands. Ecol. Appl. 16, 503–514 (2006).
Taleei, A. & Shaabani, J. Exploring genetic variation based on drought-induced phenotypic alterations during reproductive stages in Desi and Kabuli types of chickpea. Research Square, doi:https://doi.org/10.21203/rs.2.21488/v1 (2020).
Aliyu, O. M., Tiamiyu, A. O., Usman, M. & Abdulkareem, Y. F. Variance components, correlation and path analyses in cowpea (Vigna unguiculata L., Walp). Journal of Crop Science and Biotechnology 25, 173–182 (2022).
Grace, J. B. & Bollen, K. A. Interpreting the results from multiple regression and structural equation models. Bulletin of the Ecological Society of America 86, 283–295 (2005).
Yan, W. & Tinker, N. A. An integrated biplot analysis system for displaying, interpreting, and exploring genotype× environment interaction. Crop Science 45, 1004–1016 (2005).
Yan, W. & Frégeau-Reid, J. Genotype by Yield*Trait (GYT) Biplot: a Novel Approach for Genotype Selection based on Multiple Traits. Scientific Reports 8, 8242, doi:10.1038/s41598-018-26688-8 (2018).
Russell, V. L. (The University of Iowa Iowa City, IA, 2022).
Olivoto, T. & Lúcio, A. D. C. metan: An R package for multi-environment trial analysis. Methods in Ecology and Evolution 11, 783–789 (2020).
Assefa, K. The dire need to support 'orphan crop' research. Appropriate Technol. 41, 8 (2014).
Chanyalew, S. et al. Significance and prospects of an orphan crop tef. Planta 250, 753–767 (2019).
Venezia, M. & Creasey Krainer, K. M. Current advancements and limitations of gene editing in orphan crops. Frontiers in Plant Science 12, 742932 (2021).
Kumar, J., Pratap, A. & Kumar, S. Phenomics in crop plants: trends, options and limitations. (2015).
Vieira, C. C. & Chen, P. The numbers game of soybean breeding in the United States. Crop Breeding and Applied Biotechnology 21 (2021).
Hemavathy, A. T., Shunmugavalli, N. & Anand, G. Genetic variability, correlation and path co-efficient studies on yield and its components in mungbean [Vigna radiata (L.) Wilezek]. Legume Research 38, 442–446 (2015).
Cokkizgin, A., Colkesen, M., Idikut, L., Ozsisli, B. & Girgel, U. Determination of relationships between yield components in bean by using path coefficient analysis. Greener journal of agricultural sciences 3, 85–89 (2013).
Yahaya, S. & Ankrumah, E. Character association and path coefficient analysis for yield components and grain yield in soybean [Glycine max (L.) Merill.]. Legume Research 40, 630–634 (2017).
Gambín, B. & Borrás, L. Resource distribution and the trade-off between seed number and seed weight: a comparison across crop species. Annals of Applied Biology 156, 91–102 (2010).
Fletcher, R. S., Mullen, J. L., Heiliger, A. & McKay, J. K. QTL analysis of root morphology, flowering time, and yield reveals trade-offs in response to drought in Brassica napus. Journal of Experimental Botany 66, 245–256 (2015).
Karahan, T. & Akgun, İ. Selection of barley (Hordeum vulgare) genotypes by GYT (genotype × yield × trait) biplot technique and its comparison with GT (genotype × trait). Applied Ecology and Environmental Research 18 (2020).
Mohammadi, R. Genotype by yield* trait biplot for genotype evaluation and trait profiles in durum wheat. CEREAL RES. COMMUN. 47, 541–551 (2019).

No competing interests reported.

Download PDF

Editorial decision: Revision requested
10 Oct, 2024
Reviews received at journal
18 Sep, 2024
Reviews received at journal
13 Sep, 2024
Reviewers agreed at journal
29 Aug, 2024
Reviewers agreed at journal
27 Aug, 2024
Reviewers agreed at journal
27 Aug, 2024
Reviewers invited by journal
27 Aug, 2024
Editor assigned by journal
26 Aug, 2024
Editor invited by journal
14 Aug, 2024
Submission checks completed at journal
12 Aug, 2024
First submitted to journal
30 Jul, 2024

You are reading this latest preprint version

Genome-wide Marker-based dissection of genetic variability for yield and yield components, and multi-trait selection in Kersting’s groundnut (Macrotyloma geocarpum)

Status:

Version 1

Abstract

Figures

Introduction

Results

Mean performance and phenotypic variation

[Please insert Tables and around here]

Data quality, variance components and genetic parameters

Structural equation modeling of yield components

Multi-trait selection

Discussion

Conclusions

Materials and Methods

Germplasm and experiment design

Data collection

Genotype data

Statistical analysis

Declarations

Competing interests Statement

Funding

Author Contribution

Acknowledgement

Data Availability

References

Additional Declarations

Status:

Version 1