Dolichos bean [Lablab purpureus (L.) Sweet; 2 n = 22] is a versatile legume cultivated across tropical and subtropical regions of Asia, Africa and the Americas (Mohapatra et al., 2025). The crop is multipurpose, serving as a vegetable, pulse, fodder, green manure and cover crop and also possesses ornamental and therapeutic value (Raghu et al., 2018). In South India, it is an important dietary protein source, with pods and seeds rich in phytochemicals, vitamins, minerals and essential amino acids (Naeem et al., 2020). In India, dolichos bean occupies about 0.085 million hectares, largely in southern states, with Karnataka contributing nearly 90% of the area and production (Praneetha and Srivastava, 2022).
       
Numerous landraces of vegetable-type Dolichos (var. typicus) are still cultivated, characterized by photoperiod sensitivity, perennial twining (pole-type) growth habit and tender, low-fiber pods (Ramesh and Byregowda, 2016). In parallel, several photoperiod-insensitive improved varieties with determinate (bush) or indeterminate (pole) growth habits have been developed and widely adopted (Raghu et al., 2018). While genetic diversity has been extensively studied in landraces and regional germplasm (Kumar et al., 2021; Kiran et al., 2024; Shubha et al., 2024), comprehensive assessments of improved vegetable-type Dolichos remain scarce. Moreover, most previous investigations have pooled genotypes without contrasting growth habits (bush vs. pole) or usage types (vegetable vs. pulse) (Geetha and Divya, 2021; Mohapatra et al., 2025). Comparative studies that attempted such analyses were constrained by narrow germplasm bases and limited analytical depth (Das et al., 2015; Sonali et al., 2015).
       
This study provides the first detailed comparison of improved bush and pole varieties of vegetable-type Dolichos in India. It assesses genetic variability, trait associations and divergence to elucidate trait patterns that can guide parental selection, thereby supporting transgressive breeding strategies for enhancing yield potential, adaptability and trait integration in vegetable-type Dolichos.

Plant materials
 
A total of 42 vegetable-type Dolichos bean genotypes were evaluated, comprising 26 improved pole-type varieties, 9 improved bush-type varieties and 7 advanced determinate lines with vegetable-type pods (Table 1; Fig 1).




 
Experimental site and design
 
Field experiments were conducted at ICAR-Indian Institute of Horticultural Research (IIHR), Bengaluru, India (11°07′N, 77°29′E; 709 m AMSL) during the Rabi seasons of 2022-2023 and 2023-2024. Bush and pole-type genotypes were evaluated in two parallel, independently randomized experiments, each laid out in a randomized complete block design (RCBD) with three replications. Each plot comprised four rows of 6 m per genotype per replication. Bush-types were planted at 60 × 20 cm spacing, while pole-types were planted at 150 × 75 cm spacing and trained on vertical trellises. Recommended crop management practices were followed (Raghu, 2018).
 
Data collection
 
Observations were recorded on 10 randomly selected and tagged plants per replication. Data were collected on fifteen traits, grouped into phenological (days to 50% flowering, days to pod set, days to first pod harvest), vegetative (Plant height, number of primary branches, racemes per plant, raceme length), reproductive (flowers per raceme, pod length, pods per plant, 10-pod weight, seeds per pod, pod width, pod yield per plant) and nutritional (protein content) attributes.
       
For protein estimation, immature pods were dried at 60°C for 48 h and ground into fine powder. Nitrogen content was determined using a colorimetric method with Nessler’s reagent and protein percentage was calculated as nitrogen × 6.25 (dry weight basis) following AOAC (1960).
 
Statistical analysis 
 
Pooled data across years were analysed using two-way ANOVA and treatment means compared by duncan’s multiple range test (DMRT) at p<0.05. Genetic parameters such as genotypic and phenotypic coefficients of variation (GCV, PCV), genetic advance as percent of mean (GAM) and broad-sense heritability (H²), were estimated with the Variability package in R (v4.4.2). Genotypic and phenotypic correlations and path coefficient analysis (with pod yield per plant as the dependent variable) were also conducted in R.
       
Multivariate analyses included principal component analysis (PCA), agglomerative hierarchical clustering (AHC) and Pearson correlation analysis based on trait means. PCA was performed with ggplot2 and factoextra, clustering used Ward’s method with squared Euclidean distances via facto Mine R and stats and inter-trait relationships were visualized using correlation heatmaps (Corrplot, performance analytics).

Genetic variability and mean performance
 
The combined ANOVA revealed highly significant (p<0.01) differences among genotypes for all fifteen quantitative traits in both bush- and pole-type dolichos (Table 2), confirming substantial genetic variability. Year effects were significant for all traits, with pod width, pod length and seeds per pod consistently significant in both growth habits. Genotype × year interactions were significant for 11 traits in pole-types and 7 in bush-types, with raceme length, pod width, pod length, 10-pod weight, protein content and pod yield per plant commonly affected, indicating differential responses across seasons.


       
Mean values and trait ranges (Table 3) highlighted the superiority of pole-type genotypes for plant height (160.20-352.50 cm), flowers per cluster (28.00-113.00), pods per plant (55.00-575.00), 10-pod weight (45.00-277.00 g) and pod yield per plant (309.0-5000.0 g). In contrast, bush-type genotypes recorded higher protein content (20.94%) than pole-types (16.39%), suggesting a nutritional advantage.


       
Genetic parameters (Table 3) showed that in bush-types, primary branches, pods per plant and pod yield and in pole-types, pods per plant, 10-pod weight, pod yield and raceme length exhibited high PCV, GCV, heritability and genetic advance, suggesting additive gene action and good prospects for selection. In contrast, days to 50% flowering and seeds per pod showed low heritability and genetic gain across both growth habits, indicating strong environmental influence and limited scope for direct selection.
 
Trait relationships and their contribution to pod yield
 
Correlation analysis (Table 4) revealed contrasting yield-trait associations between growth habits. In bush-types, yield correlated strongly with 10-pod weight (r^g = 0.99), pod width (r^g = 0.80), number of primary branches (r^g = 0.66), days to pod set (r^g = 0.64) and pod length (r^g = 0.61). In pole-types, yield was mainly driven by the number of pods per plant (r^g = 0.60) and earliness, as reflected by days to first pod harvest (r^g = 0.99). Negative associations also differed: bush-types showed a negative correlation with days to 50% flowering (r^g = -0.74), while pole-types showed one with days to pod set (r^g = -0.41), suggesting yield advantages from early flowering in bush-types and early pod set with extended harvest in pole-types.


       
Path analysis clarified these relationships (Table 4). In bush-types, pod length (0.56), days to pod set (0.50) and 10-pod weight (0.29) exerted the strongest direct positive effects on yield (Table 4). In pole-types, 10-pod weight (2.13), number of primary branches (1.32) and raceme length (0.40) were the leading positive contributors, whereas plant height (-1.25), pod width (-1.16) and days to pod set (-0.64) showed strong negative effects, traits that, in contrast, positively influenced yield in bush-types (Table 4).
 
Multivariate analysis
 
PCA analysis
 
The first five principal components (PCs) explained 82.41% of the total variance, confirming effective dimensionality reduction. Trait loadings indicated meaningful groupings, with plant height, floral traits, raceme length, earliness and pod morphology as key contributors to population divergence. Vector loadings highlighted genotypic effects: IIHR/BD/2020-6 and IIHR/BD/2020-5 had high PC1 scores (≥4.0), influencing earliness and growth habit, while IIHR/BD/2020-1 and IIHR/BD/2020-7 showed consistently high scores across the first four PCs, reflecting broad multivariate contributions.
       
The PC1-PC2 biplot, explaining 62.22% of the variance (Fig 2), clearly separated bush- and pole-types. Pole types clustered with pod width, pod length and 10-pod weight, aligning with higher yield potential, whereas bush types grouped with protein content, raceme length and flower clusters, emphasizing quality and floral traits.


 
Cluster analysis
 
Hierarchical clustering revealed distinct trait-genotype associations (Fig 3). Fifteen traits grouped into two major clusters: A1 (seeds per pod, flowers per cluster, raceme length, protein content), linked mainly with bush-type genotypes (Cluster B2-2); and A2 (Phenology, yield, pod morphology), associated with pole types (Clusters B1, B2-1). Within B1, seven pole-types including Deepali, JDL-37, JDL-79, Arka Prasidhi, Arka Vistar, Arka Adarsh and Arka Bhavani showed high A2 but low A1 values, indicating superior yield and pod traits and marked divergence from the bush-dominated B2-2.


       
This study examined the genetic variability and trait interrelationships within improved bush and pole-type dolichos vegetables, uncovering specific patterns of trait contributions to yield and adaptability. The significant genotypic variation across all traits confirms broad genetic variability in vegetable dolichos, supporting its potential for targeted improvement. As a legume with a long history of domestication in India, Dolichos harbours extensive diversity in pod morphology, flowering behaviour and growth habit, which has been shaped by centuries of farmer-led selection for yield, quality and adaptation (Maass, 2016; Deepana et al., 2025; Mohapatra et al., 2025).
       
In the present study, pole-types excelled in plant height, pods per plant, 10-pod weight and pod yield, while bush-types surpassed them in protein content. This agrees with earlier reports that bush-types are generally preferred for nutritional quality (Sonali et al., 2015), while pole-types dominate in yield attributes due to their extended podding phase and climbing habit (Das et al., 2015; Kalpana et al., 2024; Mugali et al., 2024). Notably, both growth forms exhibited variability in protein content, with bush genotypes like Arka Amogh and IIHR/BD/2020-5 and pole genotypes such as RND-1 and GJIB-2 achieving protein levels exceeding 20%. This indicates that simultaneous enhancement of yield and quality is achievable through strategic breeding efforts.
       
Genetic parameter estimates provided insights into the underlying inheritance of traits. Pod yield and its key components exhibited high heritability coupled with substantial genetic advance in both growth habits, indicating predominant additive gene action and strong potential for direct phenotypic selection. Notably, Das et al., (2015) reported contrasting results for pod yield, with low heritability in bush-types and high in pole-types, although yield-contributing traits generally showed high heritability and genetic advance across growth habits. In contrast, days to 50% flowering and seeds per pod exhibited low heritability and genetic gain, suggesting a strong environmental influence on these traits. This is consistent with earlier studies highlighting the roles of temperature and photoperiod in regulating flowering and seed set in Dolichos (Basanagouda et al., 2022; Kiran et al., 2024; Venkatesan et al., 2024). Collectively, these findings underscore the need to identify stable yield-associated traits across environments, particularly in pole-types, which exhibited stronger genotype × season interactions. 
       
Contrasting correlation patterns between growth habits indicated different strategies for yield improvement. In bush-types, pod yield was positively associated with 10-pod weight, pod width, primary branches and pod length, highlighting the importance of pod size and branching. Negative correlation with days to 50% flowering further suggests that early flowering enhances yield in bush-types, a finding consistent with previous reports on short-duration legumes where synchrony in flowering and pod set contributes to higher yield efficiency (Egli, 2005; Mondal et al., 2011). In pole-types, yield was driven primarily by pods per plant and earliness (days to first pod harvest), reflecting the advantage of high pod load and prolonged harvest. Similar associations between pod number and cumulative yield have been reported in perennial Dolichos and cowpea (Chattopadhyay and Dutta, 2010; Gangadhara et al., 2023; Gangadhara et al., 2024). 
       
Path coefficient analysis provided deeper insights into the relative importance of traits. In bush-types, pod length, days to pod set and 10-pod weight exerted the strongest direct effects on yield, with branching and pod width contributing indirectly. This suggests that simultaneous improvement of pod morphology and early reproductive development can drive yield improvement in bush-types. In pole-types, 10-pod weight, primary branches and raceme length exerted strong positive direct effects, while plant height, pod width and days to pod set exerted strong negative effects. These contrasting direct-indirect effects highlight that traits favourable in one growth habit may constrain yield in the other, underscoring the need for habit-specific breeding strategies. Such divergence in trait-yield relationships has been documented in legumes like pigeonpea and French bean, where tall or indeterminate plants sometimes allocate assimilates away from reproductive sinks (Beebe et al., 2013; Pawar et al., 2022).
       
Multivariate analyses offered a broader perspective on trait interactions and genetic divergence. PCA revealed that a small number of traits, including plant height, floral traits, raceme length, pod morphology and earliness, accounted for most of the population divergence, confirming their utility as key discriminating traits. The PC1-PC2 biplot clearly distinguished bush and pole-types, with pole-types clustering around yield and pod traits, while bush-types clustered around protein content and floral traits. This separation aligns with earlier reports where PCA effectively differentiated Dolichos accessions based on growth habit, pod characteristics and nutritional traits (Hadavani et al., 2018; Kumari et al., 2022; Kiran et al., 2024; Shubha et al., 2024; Mohapatra et al., 2025).
       
Cluster analysis further clarified genotypic divergence. Bush-types predominantly grouped with floral and protein traits, while pole-types aligned with yield and pod traits, consistent with their contrasting breeding values (Das et al., 2015). The phylogenetic tree placed most bush-types into one cluster, while pole-types dispersed into several distinct clusters, with Cluster A5 showing maximum divergence from bush-types. This separation indicates that bush and pole-types represent distinct genetic pools, offering opportunities for complementary hybridization to exploit heterosis (Geetha and Divya, 2021; Mohapatra et al., 2025). Importantly, a genetically distinct pole cluster comprising Deepali, JDL-37, JDL-79, Arka Prasidhi, Arka Vistar, Arka Adarsh and Arka Bhavani emerged as a promising heterotic pool.

This study provides one of the most integrative assessments of genetic variability in improved vegetable-type Dolichos in India. Growth-habit-specific determinants were identified: bush-types excelled in early maturity, floral traits and protein content, while pole-types dominated in pods per plant, harvest duration and yield. Multivariate and phylogenetic analyses revealed distinct clusters, notably a pole-specific group with strong heterotic potential. By clarifying divergent yield pathways and identifying complementary genotypes, these findings offer a roadmap for targeted parental selection, enabling development of early-maturing, high-yielding and nutritionally superior cultivars adapted to diverse Indian agroecologies.

Authors gratefully acknowledge Director, ICAR-IIHR, Bengaluru, for providing the necessary infrastructure and support.

Authors state that none of the work described in this research could have been influenced by any known competing financial interests or personal relationships.