Various reports detailed assessment of phenotypic diversity (evaluated at least in two environments) following morpho-agronomic descriptors to identify diverse accessions with agronomically beneficial traits in nine underutilized grain legume species (Table 3). Analysis of 169 African yam bean (AYB) accessions involving 31 phenotypic descriptors revealed significant variation for flowering, leaf area, seeds pod^-1, pod length, seed thickness, and seed weight (Shitta et al., 2021). A detailed morphological characterization of 196 AYB landraces revealed ample genetic variation for flowering, pod and seed size, and seed yield, grouped the accessions into five clusters, and found accessions for earliness, better phenological appeal, and high seed yield. The positive and significant correlation among these traits indicates the possibility of simultaneous improvement (Olomitutu et al., 2022a). Characterization of 40 AYB accessions using 48 descriptors revealed a sizeable proportion of AYB accessions (42%) that produced tubers and four seed shapes were detected. Sixteen significant reproductive traits grouped the accessions into five distinct clusters, and found pods plant^-1, and total seed weight plant^-1 as yield determining factors (Ojuederie et al., 2015).

The genus Vigna contains many underutilized pulses (adzuki bean, Bambara groundnut, moth bean, rice bean and others) in addition to some major grain legumes (cowpea and mungbean). Adzuki bean in widely grown across China in a wide range of agroecological environments. A comparison of 475 adzuki bean germplasms including cultigens, landraces and wild relatives revealed significant variation among accessions of different germplasm types, and grouped the accessions into five clusters (Hu et al., 2022). Landraces from the lower Yellow River basin of mid-north China provinces had the greatest diversity, with various levels of cohesiveness (based on phenology, yield and yield components, and plant height data). South China (Sichuan-Anhui) accessions were typically late maturity with low seed weight, while those from north China (Liaoning-Heilongjiang) were early and had a short habit (Redden et al., 2009). Further multisite evaluation of Chinese adzuki bean germplasm showed that late maturing gene pool had the greatest yield at the lower latitude location (Hermitage, Queensland, Australia), while the central Chinese gene pool combined both high yield and acceptable seed quality, thus being suitable for marketing in Japan (Redden et al., 2012). Wild adzuki bean accessions are an excellent source of variation for root system architecture and morphological diversification to enhance the productivity of cultivated adzuki beans. High throughput root imaging analysis of 61 wild adzuki bean accessions showed wide variation (up to 17-fold among contrasting accessions) for root morphological and root architectural traits. Total root length varied from 82 to 1,435 cm, surface area from 12.30 to 208.39 cm², and average diameter from 0.23 to 0.56 mm, whereas root architectural traits, number of tips plant^-1, link average length (cm), and link average diameter (mm) were 04.33–2549.20, 0.06–0.29, and 0.27–0.61, respectively (Tayade et al., 2022a). Highly significant differences for morpho-agronomic traits among Bambara groundnut landraces and positive correlation among yield and yield attributing traits suggest that landraces could be further improved for agronomic traits (Khan et al., 2021a; Uba et al., 2023).

Hierarchical cluster analysis using 19 morphological descriptors grouped 297 Kersting’s groundnuts from across diverse ecological zones in Benin and Togo, into four clusters (Akohoue et al., 2019). Coulibaly et al. (2022a) integrated ecological niche modeling (ENM) and genetic information to understand the current and future distributions of Kersting’s groundnut populations. Both climatic and soil variables influenced the distribution of Kersting’s groundnut. #Niches projections show divergence in the response of the species and subpopulations to ongoing climate change. Thus, inclusion of genetic information into ENM may help understand species future distribution and adaptation for identifying priority regions for conservation and breeding.

In lablab, wide variation was reported among 277 accessions for pod curvature (from curved to straight), pubescence (glabrous to pubescent), fragrance (absent to high), constriction, colour, attachment, and pod colour at physiological maturity. Diverse seed colours and shapes were noted. Cluster analysis based on 14 quantitative traits grouped the accessions into four groups, with weak association between group membership and the place of origin. Variation between clusters meant that, for example, one cluster had more locules pod^-1 and seeds pod^-1, and another had a greater number of pods and yield (Letting et al., 2022). Gene pools with 2 seeds per pod and 4 or more seeds per pod are the result of independent domestication events (Njaci et al., 2023).

Arslan et al. (2022) noted significant variability for quantitative and qualitative traits among 94 grass pea accessions evaluated at low- and high-land Turkey environments. Accessions with low β-N-oxalyl-L-α, β-diaminopropionic acid (β-ODAP) content, a neurotoxin compound, were late in flowering and produced low biomass and seed yields. Resistance to broomrape (Orobanche spp.) was investigated in grass peas because this weed is a root holoparasitic plant that cause substantial loss to pulses production in the Mediterranean and sub-Saharan Africa. Assessment of wild grass pea accessions against two common broomrape species (O. foetida, O. crenata) detected complete resistance to O. crenata and O. foetida in Lathyrus articulatus and moderate resistance in L. aphaca and L. ochrus (Abdallah et al., 2021).

Multivariate analysis based on phenotypic descriptors grouped 48 Brazilian lima bean landraces into three clusters. UFPI-667 and UFPI-682 are genetically distinct and complementary in their characteristics, shorter cycle or high yield, which may be intercrossed to developed new populations (Assunçäo Filho et al., 2022). Thirty-one lima bean germplasm showed substantial variation in 100-seed weight (24–72 g) and protein content (20%–30%). Tannin content varied tenfold among the accessions that increased with seed color (Offei et al., 2006). A few lima bean landraces from the Yucatan peninsula (Mexico) when evaluated for leaf and physiological traits, herbivory insect damage and seed yield, were noticed as the best performing landraces, combining lowest cumulative herbivorous damage and high seed yield. Morphological (leaf number, area, dry mass of leaves; trichome density, specific leaf thickness, hardness) and physiological (photosynthesis rate, stomatal conductance, intercellular carbon, water use efficiency, transpiration) characteristics positively correlated with low insect damage and high seed yield (Ruiz-Santiago et al., 2021).

Tepary bean accessions (six cultivated and 19 wild among 302 accessions evaluated) grown under high temperature and acid soil conditions with aluminum toxicity produced more pods plant^-1, larger seeds, and a greater number of seeds pod^-1. These accessions revealed significant differences in physiological traits, such as flowering and maturity, specific leaf area, stomatal density, and root biomass. In addition, cultivated accessions had higher photochemical quenching (qP), while energy dissipation by non-photochemical quenching (NPQ) in the form of heat and the coefficient of non-photochemical dissipation (qN) were higher in cultivated regressive and wild accessions, which probably contributed to differences in adaptation to combined stress of high temperature and acidic soil conditions. Six and 19 accessions of cultivated and wild groups, respectively, had grain yields above 1.8 t ha^-1, which may be deployed in breeding program to improve productivity of tepary beans in such environments (Suárez et al., 2022).

The daylength-neutral accession MY0-01 from Bago (Myanmar) is adapted to southern Australia and is an excellent genetic resource to breed photoperiod sensitive winged bean cultivars. The small pod and hard-seededness, respectively, may lower its potential for vegetable production and adversely affect germination and plant establishment (Eagleton, 2022).

Table 4 lists DNA marker-based assessment of population structure and diversity among select germplasm to identify genetically diverse accessions in 13 underutilized grain legume species.

Single nucleotide polymorphism (SNP)-based diversity assessment involving 169 AYB grouped the accessions into three subpopulations with high genetic differentiation. Subpopulation 1 accessions were high yielding, while those in those in subpopulation 2 were highly polymorphic and heterozygous (Shitta et al., 2022). Another study involving 93 AYB accessions and 3722 SNPs and multiple clustering methods detected substantial genetic diversity and formed three to four clusters, with most accessions in each cluster having a similar phenotype, i.e., seed or seed and tuber types (Aina et al., 2021). Insertion-deletion (InDel) markers differentiated salt tolerant from sensitive AYB accessions (Zhang et al., 2021).

A study involving 261 adzuki bean accessions from China and 163 simple sequence repeats (SSRs) grouped the accessions into 10 clusters. Accessions from northern China were genetically distinct than those from southern China (Chen et al., 2015). Wild adzuki beans had higher allelic diversity than cultivated types. Structure analysis clearly separated the wild from the cultigen germplasm, with subdivisions in cultigens based on ecological regions of adaptation (Wang et al., 2012). High genetic differentiation was reported between wild adzuki beans and their wild relative Vigna minima. Accessions could be distinguished from each other based on their origins, which suggests that geographic regions of adaptation of wild adzuki bean shaped their genetic variation. Further unfolding of relationship between Chinese adzuki bean cultivars and wild adzuki bean accessions highlights their closeness to Japanese wild adzuki beans than to domestic accessions, suggesting greater involvement of Japanese adzuki bean accessions in Chinese adzuki bean breeding (Liu et al., 2014).

Profiling of 93 South African Bambara groundnut accessions using 2,286 SNPs and morpho-agronomic traits, revealed moderate genetic differentiation and two distinct clusters (Majola et al., 2022). Highly polymorphic SNPs (5927) differentiated 100 Nigerian Bambara groundnut accessions into seven subpopulations and significant marker-trait associations (MTAs) for several morpho-agronomic traits (Osundare et al., 2022). DArT SNP profiling of 270 Bambara groundnut landraces revealed three subpopulations which corresponded to geography. Accessions from West Africa and of unknown origin formed subpopulation 1, Central Africa accessions subpopulation 2, and those from southern and eastern Africa subpopulation 3 (Uba et al., 2021).

Validation and diversity analysis involving 33 polymorphic SSRs and 58 horse gram germplasm detected wide variation and grouped the accessions into 3 (Structure analysis) or 4 (UPGMA and PCA) distinct clusters. The early flowering types are clearly separated from late flowering group (Kumar R. et al., 2020). Another study involving 48 horse gram germplasm and 117 SSRs grouped the accessions into two distinct groups (Kaldate et al., 2017). High quality SNP (493) data on 281 Kersting’s groundnut accessions formed four clusters (based on a neighbor joining tree), which were differentiated by seed coat colour, while structure analysis, yielded two subpopulations. Most of the accessions from the Sudan savanna were in subpopulation I, while those from Sudano-Guinean and the Guinean savannas were in subpopulation II. This study also detected 10 significant MTAs, of which six SNPs were consistent across environment (Akohoue et al., 2020).

Profiling of 25,039 DNA markers (9320 DArT SNPs and 15,719 SilicoDart) data on 65 lablab bean germplasm showed low discriminating ability and three subpopulations unrelated to site of origin. High within population variance suggests a greater degree of gene exchange or low genetic differentiation among the populations (Sserumaga et al., 2021). Another data set involving 1,000 SNP and SilicoDArT markers data on 142 lablab accessions unfolded five major groups, each with further subgroups (Muktar et al., 2021).

The presence of β-N-oxalyl-L-a,b-diaminopropionic acid (β-ODAP) causes neurological disorders in humans. Analysis of 56 SSRs on 400 accessions of grass pea germplasm showed highly diverse structure, with two main and one admixed population (Rahman et al., 2022). Analysis of 31 EST-SSRs data on 22 grass pea accessions having low β-ODAP content detected two main clusters, with high genetic distance between some pairs of accessions, which may be used in crop improvement program to develop populations with large variation to pursue selection for low β-ODAP content (Arslan et al., 2020). Eighteen EST-SSR data from 118 accessions including wild relatives formed four clusters. High β-ODAP accessions (mostly wild types) were in cluster I and those with low β-ODAP in cluster II (Gupta et al., 2018). A large study using 30 SSRs grouped 283 grass pea accessions into three clusters, i.e., wild species, Asian accessions, and Europe and Africa accessions. Asian accessions were clearly separated from other groups (Wang et al., 2015).

Lima bean germplasm was extensively analyzed for assessing population structure and diversity. A combined analysis of genotyping (12 SSRs) and phenotyping (7 morphological descriptors) data of 183 Brazilian lima bean landraces detected high diversity and three distinct subpopulations, one predominantly from the Andean gene pool with large seeds (mean 100-seed weight 80 g) and the other two predominantly from the Mesoamerican gene pool (mean 100-seed weight 34 g), with considerable introgression between the Andean and the Mesoamerican gene pools (Silva et al., 2019). Analysis of 73 inter simple sequence repeats (ISSRs) on 46 Mayan lima bean landraces detected high levels of diversity and genetic differentiation. Mayan low landraces showed higher genetic diversity than Mayan high landraces, thereby indicating the influence of Mayan culture on diversification and conservation of lima beans (Camacho-Pérez et al., 2018). Wild lima beans have two gene pools, known as MI and MII. Genomic profiling of 67 wild populations of lima bean from Mexico with 10 SSR markers, however, unfolded three gene pools, MI1a, MI1b, and MII, with greater among population than within populations variance. MI and MII were geographically divergent, while MI1a and MI1b overlap in their distribution and the presence of admix individuals suggests geneflow among gene pools (Martínez-Castillo et al., 2014).

Molecular profiling of 472 rice bean accessions (388 cultivated and 84 wild) detected high gene diversity in cultivated populations, which was ∼83% of that for wild populations. East Asian populations formed a distinct gene pool. The cultivars from Indonesia had a genetic structure like the wild accessions. Accessions from western Nepal were quite distinct from others and formed a specific group, thereby being a unique gene source for rice bean breeding (Tian et al., 2013). Northeastern Himalayan hills of India are the hotspot for biological diversity. The rice bean populations from this region showed exceptionally high outcrossing rate. Characterization of 65 rice bean accessions from this region with 28 SSRs revealed high mean gene diversity bordering eastern Nepal and Myanmar, formed three distinct clusters, with accessions bordering Myanmar and eastern Nepal in a distinct cluster (Iangrai et al., 2017).

Molecular profiling of 158 cultivated and wild tepary bean populations with high quality SNPs (768) separated domesticated and cultivated tepary beans, with two distinct groups within the domesticated types (Gujaria-Verma et al., 2016). Furthermore, characterization of 20 cultivated accessions with ten highly polymorphic SSRs revealed moderate differentiation among genotypes. SSR-based diversity correlated with diversity assessment based on morphological descriptors, and a few genetically distinct accessions for use in breeding (Mhlaba et al., 2018).

Molecular profiling of 457 accessions winged bean accessions with 14 SSRs revealed moderate gene diversity and high genetic admixture. Seed exchange and relatively high outcrossing probably contributed to high genetic admixture in this germplasm set. Structure analysis grouped the accessions into three subpopulations, unrelated to geographic origins, with most accessions having long pods (30 cm or greater in length), purple seed coats or young purple pods grouped together (Laosatit et al., 2022).

Overall, assessment of population structure and diversity among gene pools provides greater insights into the genetic makeup of populations, which will help in conservation strategy, management, and utilization of diversity in crops breeding and genetics. Unlike major pulses where there has been extensive use of genetic resources in crop breeding, the diversity in underutilized pulses remains largely untapped as a source of alleles for introgression. Hence the primary gene pool can harbor unexploited genes for crop improvement.

The declining nutritional quality and bioavailability of nutrients results in serious malnutrition (Dwivedi et al., 2013; Owino et al., 2022). Underutilized pulse grains are rich sources of seed-protein, minerals, vitamins, and phytonutrients, and vary in hydration, cooking, textural and pasting properties. For example, seed weight and protein content among adzuki bean accessions, respectively, varied from 75 to 148 g 1,000 seeds^-1 and 19%–24%. The accessions also showed significant variation in their physicochemical, cooking hydration and textural properties. Soaked grain hardness ranged between 69 and 120 N and had significant positive association with gumminess and chewiness (Yadav et al., 2018). Assessment of physicochemical and digestive properties of starches from Japanese “dainagon” adzuki bean variants revealed larger particle size in “Noto-dainagon” and “Kyoto-dainagon” starch types were easily gelatinized, none of their starch showed breakdown, and ‘dainagon’ starch was more indigestible than others (Honda et al., 2020). Adzuki bean seedcoats are rich in phenolics and antioxidant compounds, thus having a potential health benefit to humans (Johnson et al., 2022). Total metabolites, the saponin and phenolics contents, ranged from 16 to 945 mg DE g^-1 and 0.80–57.35 mg GAE g^-1, respectively. Delphinidin-3-O-glucoside and delphinidin-3-O-galactoside were the predominant anthocyanins in black-seeded adzuki bean cultivars (Desta et al., 2022). Seed coat is a rich source of metabolites and closely associated with the anthocyanin and flavonoid metabolism pathways. Analysis of 10 seed coat colour types in adzuki bean including red, black mottle or gray, golden, green, black, ivory, brown, and light brown revealed anthocyanins the main pigment source, with no carotenoid or pelargonidin derivatives in the seed coats. The pigment composition of the different seed coat colours and the combination of proanthocyanidins and anthocyanins affect the seed coat colour in adzuki bean (Zhao et al., 2022).

Bambara groundnut landraces showed considerable differences in total phenolics (0.75–17.71 mg GAE g^-1), flavonoids (0.01–2.51 mg QUE g^-1) and anthocyanins (0.03–1.31 mg CYE g^-1), with caffeic and catechin discriminating the landraces (Tsamo et al., 2018), while a comparison of phenolic in whole and dehulled Bambara groundnut cultivars differing in seed colour varied from 3.6 to 11.0 GAE g^-1 and from 2.7 to 3.2 GAE g^-1, respectively (Adedayo et al., 2021). A new compound, luteolin (C₂₀H₁₈O₉), with antioxidant activity discovered in Bambara groundnut suggests that seed could be used as natural sources of antioxidants to reduce inflammation in humans (Chinnapun and Sakorn, 2022).

The seed protein content of grass pea accessions belonging to both L. sativus and L. cicera ranged between 25% and 30%. The protein is of high quality, except for having low methionine. The seeds are not rich in fat (<1%) but contain high level of polyunsaturated fatty acids. L sativus accessions had relatively low level of β-ODAP content, average 733 mg with a range of 0.583–1.340 mg kg^-1 DM, while those from L. cicero accessions had slightly higher β-ODAP, average 1,168 mg with a range of 0.911–1.349 mg kg^-1 DM. L. cicera seeds on average contain greater (6.4 g kg^-1 DM) tannins than L. sativus seeds, 3.3 g kg^-1 DM. The tannin content in L. sativus seeds were significantly correlated with flowers and seeds colour, darker colour associated with higher levels of tannins (Grela et al., 2012). Assessment of seed protein and β-ODAP contents among 702 L. sativus accessions revealed significant differences in seed protein (28.82%–30.72%) and β-ODAP (0.32%–0.47% and 0.38%–0.53%), and reported accessions low in seed protein and β-ODAP contents (Rajendran et al., 2019). The landraces from Turkey showed greater variability in seed protein (24.07–30.9%) and β-ODAP (1.35–3.86 mg g^-1 DM) contents, with a few identified as low in seed β-ODAP and high protein contents (Basaran et al., 2011).

The protein content among 96 diverse horse gram germplasm ranged from 13% to 40%, with the highest protein noted in the wild species Macrotyloma sar-gharwalensis. The number of metabolites varied from 25 to 44, and the most nutritionally diverse germplasm amongst the panel were IC 280031 and IC 139356, with the greatest number of quantifiable metabolites, which may be used for the development of nutraceutical food for humans. Chemo-markers such as methionine, sucrose, maltose, riboflavin, and myricetin effectively differentiated this panel (Gautam and Chahota, 2022). Assessing the chemo-diversity of horse gram germplasm cultivated for eight seasons in the Alpine Himalayas unfolded 46 diverse metabolites, including 18 amino acids, nine carbohydrates, three vitamins, seven flavonoids, and nine miscellaneous molecules, which generated a single snapshot chemical signature to differentiate accessions. Higher elevation (1829 m asl) seed-produce resulted in greater amounts of metabolites (55.78 g kg^-1 DM), than those obtained from lower (1,000–1,150 and 1,150–1,450 m asl). Chemotype-based PCA formed three distinct clusters, with cluster 3 accessions containing the nutritionally best metabolites (Gautam et al., 2022). Horse gram seed is a good source of bioactive and nutritive compounds compared to other pulses and has a high level of antioxidant and radical scavenging activities with immense health benefits (Ingle et al., 2021).

Variation in fatty acid profiling of lablab bean suggests its grain is the good source of essential omega-6 fatty acid linoleic acid (C18:3, ω-6) and omega-3 linolenic acid (C18:3, ω-3), in addition to containing high amounts (24%) of proteins and minerals but low (1%) in phytate content (Hossain et al., 2016; Kumari M. et al., 2022). Low phytate content in the seed enhances mineral bioavailability, which may mitigate nutritional vulnerability in the developing world (Dwivedi and Ortiz, 2021). Significant variability in phytochemicals: i.e., trypsin inhibitors (0.041–0.078 TIU mg^-1), phytic acid (0.26–1.22 mg g^-1 DM), radical scavenging activity (5.63–38.26), and tannins (1.4–3.2 mg g-1 DM), were reported among moth bean germplasm (Gupta et al., 2016).

Bhagyawant et al. (2019) noted significant variation in seed protein (2.88–31.79 mg g^-1 DM) content and antioxidant activity (9.36–18.56%) among rice bean germplasm, with some accessions showing highest in vitro protein (63.84%) and starch (44.6%) digestibility. Accessions with high antioxidant activity, protein content, increased in vitro protein and starch digestibility and reduced oligosaccharides could be used as a donor parent or promote them for human consumption.

Assessment of wild and domesticated tepary bean accessions for 100-seed weight and protein contents revealed higher level of protein and lower seed weight in wild than domesticated tepary beans. However, domesticated tepary beans had a larger range variation in protein values than the wild tepary beans (Waines, 1978). Tepary bean relative to common bean showed reduced fat and ash concentration but higher sucrose content. Shorter cooking time and a high percentage of seeds with measurable water uptake were reported in tepary bean than common bean accessions, while a few lines were of ‘hard-shell’ trait (i.e., low water uptake) and longer cooking time. Tepary bean is highly nutritious, with nutrient composition and cooking characteristics like that of common bean (Porch et al., 2017). A few common bean × tepary bean interspecific congruity-backcross lines (#12, 76, 77, 78), though yielded poorly, had seed-Fe content over 100 mg kg^-1 DW (Burbano-Erazo et al., 2021). Hence greater efforts are needed to form such subsets in other underutilized pulses to enhance germplasm use in crop improvement programs.

Reduced core and mini core subsets (defined by morphological descriptors or genotyping data and representing the diversity of the entire collection of a species accessions preserved in a genebank) are the ideal set of genetic resources for detailed characterization to dissect population structure and diversity, discovering trait-specific accessions, gene discovery and allele mining. The underutilized pulses, unlike major grain legume crops, received low priority in management and utilization of genetic resources in genetic enhancement programs. To date, core collections have been reported in adzuki bean (Ning et al., 2008; Takeya et al., 2013), Bambara groundnut (Uba et al., 2023), lablab bean (Vaijayanthi et al., 2015), lima bean (Gomes et al., 2020), and moth bean (Meghwal et al., 2015).

The availability of genetic markers discussed above allows evolutionary and genomic insight into crop origins and the genetic basis of adaptive traits. These markers have revolutionised the identification of QTL; i.e., regions of the genome controlling one or more traits of interest (Paterson et al., 1988). Small numbers of genetic markers can give coarse resolution of QTL, but as more become available the genomic interval controlling a particular trait can be narrowed down, sometimes to the causative gene (reviewed in Burke et al., 2007). Research using this approach was the first to identify genes underlying important agronomic traits, e.g. the loss of shattering in domesticated rice (Konishi et al., 2006; Li et al., 2006), hard kernel coat in maize (Wang et al., 2005), and fruit size in tomato (Frary et al., 2000). High density linkage maps are available for several minor legumes, including adzuki bean (Han et al., 2005), Bambara groundnut (Gao et al., 2023), grass pea (Hao et al., 2022), lima bean (Garcia et al., 2021), and winged bean (Chankaew et al., 2022).

Even dense linkage maps may give poor resolution of causative genes if recombination is low, or the number of plants required is prohibitively large (the rice examples above both required >10,000 plants, which would be unreasonable for species where the plants are larger). However, identifying a genomic region and knowing the sequence of the entire region can remedy this, allowing the researchers to find candidate genes based on likely gene function. The sequencing, assembly and annotation of crop genomes is therefore a significant step in any analysis of agronomic traits. Once available, this can fast-track the identification of causative genes and alleles, as well as identify markers for marker-aided breeding (Jackson et al., 2011).

Whilst modern high throughput sequencing technologies can sequence a genome many times over for a fraction of the cost required for the first genome assemblies, the most accurate and complete assembly requires significant investment, aided by very long read sequencing, and large mapping populations to help unify scaffolds into chromosomes. This means that while many minor legume crops have been investigated with genome sequencing, the availability of high-quality contiguous genome sequences of minor legumes is low, with the two best examples probably being pigeonpea and chickpea, which, due to their widespread use and research investment, probably cannot be considered ‘minor’ legumes. Therefore, focussing on those in Table 2, we can see that adzuki bean (Kang et al., 2015; Yang et al., 2015), lima bean (Garcia et al., 2021), tepary bean (Moghaddam et al., 2021), horse Gram (Shirasawa et al., 2021), and lablab (Njaci et al., 2023) have genomes available with a good contiguity and coverage, and grass pea has a recent attempt to assemble this very large genome (Rajarammohan et al., 2023). Several more are in progress (see Chapman et al., 2022), for example Bambara groundnut (Salazar-Licea et al., 2022), and winged bean (Tsoutsoura et al., 2022).

For any species it is important to remember that a reference genome, even when contiguity and coverage are very high, only represents one individual, and there are expected to be genomic regions in other individuals that are absent from a single reference. This relates to the concept of pangenomes, wherein a complete species’ genome can only be estimated when a range of accessions are investigated (Della Coletta et al., 2021; Chapman et al., 2022). Genomic regions present in only some accessions are termed ‘dispensable’ and are typically highly repetitive and with a low density of coding regions (Gao et al., 2019). Despite this, they can contain enriched gene ontology categories of important gene types, for example defense-related genes in Brachypodium (Gordon et al., 2017) and an important flavour-related gene in tomato (Gao et al., 2019).

Whilst a highly complete and contiguous genome is desirable, the cost and resources required can be prohibitive. A low coverage and fragmented genome can be generated for a fraction of the cost, for example only using Illumina-based short-read sequencing. Despite repetitive regions of the genome prevent the assembly of larger fragments, this fragmented assembly can be gene rich and used to identify thousands of molecular markers and complete coding regions that are useful for investigations into the crop diversity and variability and understanding coding sequence-based selection across species during their evolution (Fisher et al., 2022).

Genetic markers are useful for understanding the partitioning of genetic variation among crop accessions as well as identifying closely related wild taxa that may be useful for breeding. Genetic markers, often SSRs and SNPs have been used in a variety of minor legumes to understand domestication, gene flow and population level variation (Table 4).

Carrying out whole genome resequencing of a broad range of germplasm and mapping this back to the reference genome, i.e., population genomics, ensures that much larger numbers of genetic markers, potentially hundreds of thousands to millions, are incorporated into the analysis. Through this, and if the sampling is appropriate, more detailed information on crop origins can be determined, for example in the underutilised crops pearl millet (Varshney et al., 2017), fonio millet (Abrouk et al., 2020) and Guinea yam (Sugihara et al., 2020). Whilst minor legumes are further understudied relative to these, recent work using population genomics has, for example, confirmed two origins of domesticated lablab (Njaci et al., 2023). A further advantage of this depth and breadth of sequencing is the ability to correlated genotype and phenotype in a genome-wide association study (GWAS). This approach has been employed in several crops, including some minor legumes too. Through this the genetic basis of reduced alkaloids in lupin (Wang et al., 2021) and flowering time in rice bean (Guan et al., 2022) have been, in part, resolved.

Genome evolution can involve indels, inversions, translocations, and fissions and fusions of chromosomes; in some cases, these can reduce or prevent gene flow. In the wild, prevention of gene flow between species is expected to, at least in part, maintain species integrity (Mallet, 2005). Within a crop (or between a crop and its wild relatives), these structural changes could potentially impede breeding attempts.

Synteny analyses demonstrate how related species’ genomes ‘stack up’ against each other. Several scientists have used genetic markers to determine the rate of translocations, fissions and fusions over species divergences, e.g. in Solanaceous crops (Wu and Tanksley, 2010). However, these only provide a coarse scale determination of these parameters and would miss small-scale inversions and indels. For more fine-scaled analyses of both synteny, a whole genome sequencing approach must be pursued. Recent work, especially in legumes, have revealed how different species’ genomes compare in terms of gene order and chromosome similarity. Sequencing of the cowpea (V. unguiculata) genome and comparison to common bean was used to enumerate a chromosome numbering system in legumes. In this comparison six of the 11 chromosomes are largely syntenic, and five of the cowpea chromosomes are formed from parts of two common bean chromosomes (Lonardi et al., 2019). These were then compared to the genomes of adzuki and mungbean, thus revealing how the genomes have evolved during the last few million years. The cowpea genome analysis also revealed that a 4.2 Mb region of the genome was inverted in some accessions relative to others, highlighting how large structural variants are found both when comparing different species and within species (Lonardi et al., 2019; Figure 1). In a mapping population wherein the parents different in the orientation of the inversion there was an absence of recombination, a well-known effect of inversions (Kirkpatrick, 2010), and within this region was a gene putatively involved in resistance to the parasitic weed, Striga spp., although the functional significance remains to be explored (Lonardi et al., 2019).

Synteny analyses have been carried out for the minor legumes lablab (Njaci et al., 2023), moth bean (Yundaeng et al., 2019) and adzuki bean (Yang et al., 2015). This research reveals how the number of translocations, fissions and fusions increased with genetic distance. An in-depth analysis of synteny between lupin and multiple other legumes allowed for the determination of an ancestral legume karyotype as well as characterizing the lupin genome as being palaeotriploid (Hufnagel et al., 2020).

Using a diverse panel of 98 wild and cultivated Vigna accessions from 13 species evaluated for agronomic traits for two seasons and genotyped with 92 cross-genera and cross-species SSRs, Kumari G. et al. (2022) detected three genetically distinct subpopulations and association of 13 SSRs with nine traits and seven markers associated with multiple traits. For example, VR022 for 100-seed weight and pod length; CEDG033 for days to flower and maturity; CEDG100 for 100-seed weight, plant height and terminal leaf length; CP1225 for chlorophyll content (CC) at 30 days, days to flower and maturity; and CEDG096A for CC30 and CC45 days, etc. CEDG100 co-localized in gene-encoding histone-lysine N-methytransferase ATX5, while VR22 co-located in gene-encoding SHOOT GRAVITROPISM 5 in mungbean, thus suggesting these markers as potential genomic resource for marker-assisted genetic enhancement of mungbean and related Vigna species (Kumari G. et al., 2022).

The introgression breeding method has been used to improve crops since domestication (Hernandez et al., 2020). It is an important method for the breeding of crops to improve their productivity and resilience under stresses caused by climate change (Gramazio et al., 2021). It involves the integration of genetic materials from one species to another, often from a wild relative to a domesticated crop (Quezada-Martinez et al., 2021). Wild crop relatives are sources of alleles for adaptation to unfavorable climatic conditions, and resistance to pathogens and pests (Hernandez et al., 2020; Gramazio et al., 2021).

Crops are domesticated through selection for desirable traits such as the selection of high-yielding cultivars under optimum conditions. However, this approach has led to a narrow genetic base that causes crop failure under climatic stress (Quezada-Martinez et al., 2021). As climate change ushers in unfavorable conditions, domesticated crops may not withstand the negative effect accompanied by these conditions. Introgression breeding can be used to transfer traits responsible for high productivity and stress tolerance from wild relatives to domesticated crop species for increased crop productivity and resilience. This method is essential for the improvement of underutilized crops by introgressing desirable traits from their wild relatives.

This method, however, is constrained by some biological factors such as the difficulty of hybridization, large genetic distance between the donor and recipient, low crossover frequency and distributions in the hybrid, and selection for desirable introgression while minimizing linkage drag can be tedious (Quezada-Martinez et al., 2021). Introgression involves the crossing of wild relatives with domesticated crop species and several backcrossing of the resulting hybrids with parents however if either the hybridization of the wild relative and domesticated crop is difficult or impossible introgression cannot take place as in the case of lima bean and its relatives’ common bean, and also with runner bean (Phaseolus coccineus) and tepary bean. The inability to hybridize could be caused by many factors, which can include large genetic distance between the wild relative and domesticated crop species. In cases where crosses were possible, the fertility of hybrids were mainly dependent on compatible chromosomal arrangement of the wild parent (Singh et al., 2018). Introgression can also be constrained by low level of gene flow in the hybrids, which leads to the non-expression of introgressed genes. Despite the challenges accompanied by introgression this method is effective in the transfer of desired traits from wild relatives to domesticated underutilized crops.

β-ODAP content varied widely (0.02%–2.59%) among Lathyrus germplasm, with greater content in stressed environments. A few wild relatives such as L. cicero, L. amphicarpus and L. ochrus have zero or low β-ODAP (≤0.01%). They may be utilized for the development of toxin free Lathyrus cultivars. Crossbreeding has resulted in the release of several high yielding grass pea cultivars with low β-ODAP content in many countries in Africa, Asia, Australia, and America. Impressive progress has been achieved towards applying biotechnological resources, including transgene technology, in the genetic enhancement of grass peas (Das et al., 2021).

Lack of rapid screening technique for identification of low β-ODAP germplasm or segregants in breeding populations is the major impediment to grass pea breeding. A recently developed high throughput plate assay based on spectrophotometric method allows quantification of total β-ODAP in a large number of samples, but its low sensitivity and inability to differentiate α- and β-L-ODAP limits its usefulness. Use of stable isotope as internal standard with a novel liquid chromatography mass spectrometry (LCMS)-based method for β-L-ODAP quantification facilitates accurate identification and characterization of grass pea lines with a very low ODAP content (Emmrich et al., 2019). An improved HILIC-MS/MS method without sample derivatization determines both toxic (β-ODAP) and nontoxic (α-ODAP) isomers in grass pea. It uses a hydrophilic interaction chromatography (HILIC) column and an isocratic gradient of eluents to determine both α- and β-ODAP contents. β-ODAP content in a validation study involving 107 geographically diverse grass pea accessions ranged between 0.45 and 6.04 mg g^-1 dry seeds, differentiated contrasting accessions, and showed moderate correlation (0.65) between α- and β-ODAP contents, reinforces independent quantification of both ODAP isomers (Bento-Silva et al., 2019).

β-ODAP synthase (BOS), a recently discovered enzyme of the benzylalcohol O-acetyltransferase, anthocyanin O-hydroxycinnamoyltransferase, anthranilate N-hydroxycinnamoyl/benzoyltransferase, deacetylvindoline 4-O-acetyltransferase superfamily of acyltransferases, provides catalytic activity linked with β-ODAP formation. It is structurally similar to hydroxycinnamoyl transferase. BOS expression in the presence of its substrates causes β-ODAP production in vivo, which may pave the way to engineer β-ODAP–free grass pea cultivars (Goldsmith et al., 2022a). Oxalic acid, a small metabolite with several metabolic pathways to control oxalate levels by enzymatic degradation, protects plants from herbivores damage. LsOCS, grass pea oxalyl CoA-synthetase gene, encodes a monomeric protein of 56 kDa, having catalytic efficiency with oxalate similar to that of Arabidopsis thaliana (AtAAE3) and Medicago truncatula (MtAAE3) homologs. Substituting LsOCS with oxalate oxidase or decarboxylase could reduce the level of β-ODAP in grass peas. Inactivating LsOCS in grass pea using genetic engineering has potential to reduce the biosynthesis of β-ODAP, however, it may increase grass pea susceptibility to pathogens such as Sclerotinia sclerotiorum. Thus, replacing LsOCS with an exogenous oxalate-oxidase or decarboxylase could enable plants to regulate cellular oxalate levels while reducing the levels of β-ODAP in grass pea (Goldsmith et al., 2022b). Germplasm with low β-ODAP may show reduced stress tolerance due to reduced relative water content and perturbed abscisic acid levels (Verma S. K. et al., 2022).

Beetles (also known as weevil, Callosobruchus maculatus) cause substantial damage to pulse grains during storage. Early research shows that Arcelin (Arl), an insecticidal gene from lablab bean, is homologous to Arl-3 and Arl-4 alleles from Phaseolus spp, with about 70% amino acid similarity. The artificial diet containing Arl (0.2% w/w arcelin-incorporated artificial seeds) retarded growth of cowpea weevil, which may be deployed to incorporate bruchid resistance through transgenesis in pulses (Sundaram et al., 2012). Exceptionally high Arl gene expression in wild sward bean (Canavalia virosa) suggests it may be used to develop weevil resistance in other cultivated pulses (Sakthivelkumar et al., 2014). An insecticidal activity of Arl gene, isolated from a wild accession of lima bean (Phaseolus lunatus), drastically reduces adult emergence and seed damage, thereby demonstrating the effectiveness against the bruchid beetle (Hilda et al., 2022).

The single dominant gene Rcc confers beetle resistance in moth bean. One major (qVacBrc2.1) and one modifying (qVacBrc5.1) QTL residing in Rcc control resistance. qVacBrc2.1, mapped on LG2 between SSRs CEDG261 and DMB-SSR160, accounted for 50%–64% variation for resistant traits. qVacBrc2.1 is the same as QTL Brc2.1, which confers beetle resistance in wild adzuki bean (Somta et al., 2018). Fine mapping revealed two novel markers associated with qVacBrc2.1 constituted two linked QTL, qVacBrc2.1-A and qVacBrc2.1-B, and two polygalacturonase-inhibiting protein genes, VacPGIP1 and VacPGIP2 as candidate genes for beetle resistance in TN67. The alignment of VacPGIP1 coding sequences between TN67 (resistant) and ICMP0056 (susceptible) accessions revealed eight SNPs, three of which altered the amino-acid sequence of the predicted domains of polygalacturonase inhibitors in ICPMO056 (Gamage et al., 2022).

Mungbean yellow mosaic India virus (MYMIV) causes significant yield losses in pulses including black gram (Vigna mungo) and green gram (Vigna radiata). High level of resistance to MYMIV was reported in a black gram variety Mash114. A large effect QTL, qMYMIV6.1.1 spanning 3.4 Mb on chromosome 6, identified as an inter-specific introgression from rice bean, accounted for 70% of total phenotypic variation. KASP markers closely associated with MYMIV delineated 500 kb genomic region linked with MYMIV, which can be deployed for marker-assisted transfer of introgressed region into improved genetic backgrounds of Vigna species (Dhaliwal et al., 2022). Advanced lines originating from a rice bean and black gram cross showed wide variation in seed yield, ranged from −35.48 to +50.31% over control (Mash338), and were found resistant to MYMIV, Cercospora leaf spot and bacterial leaf spot. The superior yield performance and disease resistance traits were introduced from the rice bean genotype KUG114, and offspring with yield superiority of ∼39% over ‘Mash338’ has been released as “Mash114” for cultivation in Punjab (Singh et al., 2013).

The genus Phaseolus contains many species, including tepary bean, a valuable genetic resource for abiotic stress tolerance and productivity genes. Wild tepary bean accession W6 15,578 is a potential donor for cold tolerance. A three-year field assessment of an interspecific backcross population derived from the cross W6 15,578 × NY5-161 (common bean) led to identifying lines tolerant to sub-zero temperature at seedling stage. Their subsequent evaluation revealed that few outyielded their common bean parent under cold stress and drought, thereby suggesting that introgression of a proportion of tepary bean genome into common bean is a promising strategy to enhance abiotic stress adaptation in the latter (Souter et al., 2017). However, obtaining hybrids between tepary and common bean is problematic (i.e., hybrid sterility) and therefore pollination technique, growth conditions and embryo rescue methods are used for successful introgression of tepary genes into common bean (Andradf-Aguilar and Jackson, 1988; Scott and Michaels, 1990; Pratt and Gordon, 1994).

Barrera et al. (2022) recycled novel interspecific derived lines obtained from common bean × tepary bean and hybridized these to Phaseolus parvifolius to increase male gametic diversity to facilitate interspecific crossing. Introgression of such lines enhanced the success of common bean and tepary bean hybridization without the use of embryo rescue technique and resulted in a 12-fold more hybrid plants than crossing directly between common beans and tepary beans. Such lines contain large introgression of genomic regions from P. parvifolius, thus providing means to efficiently exploit tepary gene pool for enhancing abiotic stress adaptation in common bean without the need for embryo rescue (Barrera et al., 2022).

A multi-environment evaluation of interspecific congruity backcross lines, obtained from common bean and tepary bean cross, resulted in a few lines that produced high yield under extreme weather conditions of coastal Colombia, thereby suggesting that it is feasible to combine drought and heat stress tolerance, as evidenced in line 68, with high Fe mineral biofortification (Burbano-Erazo et al., 2021). A few elite lines with large seeds and erect plant architecture, abiotic stress adaptation, and resistance to bacterial blight and weevil may be recycled to enhance productivity and stress tolerance traits in common bean (Porch et al., 2013; 2022).

Arabidopsis transgenics overexpressing MuHSP70, a gene cloned from horse Gram, exhibit multiple abiotic stress tolerances, which result in greater shoot biomass, root length, relative water content, and chlorophyll content during multiple stresses (Masand and Yadav, 2016). Transgenic peanuts containing MuNAC4, another gene from horse Gram, exhibited significantly enhanced drought tolerance due to increased later roots and greenish growth by reducing damage to membrane structures and enhancing osmotic adjustment and antioxidant enzyme regulation under stress (Pandurangaiah et al., 2014).

MuWRKY3 expression alone or simultaneous co-expression with other stress-responsive regulatory TFs genes (MuNAC4, MuMYB96), all cloned from horse Gram, improved drought tolerance in peanuts (Kiranmai et al., 2018). Transgenic plants showed increased growth of lateral roots, chlorophyll content, stay-green, and maintained higher relative water content compared to WT. Expression analysis of transgenes and their downstream regulatory genes revealed two-to four-fold increase in transcript levels under drought stress in multigene transgenic peanut plants over WT. Thus, multiple genes transfer with simultaneous expression is a promising option to improve stress tolerance and productivity in peanuts and possibly in other pulses in drought stressed environments.