Identification of minimum descriptors for characterization of Capsicum spp. germplasm

Identificação de descritores mínimos para caracterização de germoplasma de Capsicum spp.

RESEARCH PESQUISA

Identification of minimum descriptors for characterization of Capsicum spp. germplasm

Identificação de descritores mínimos para caracterização de germoplasma de Capsicum spp.

Waldir C Junior e Silva^I; Sabrina Isabel C de Carvalho^II; João B Duarte^I

^IUFG, Escola de Agronomia, C. Postal 131, 74001-970 Goiânia-GO; [e-mail]; [e-mail]

^IIEmbrapa Hortaliças, C. Postal 218, 70351-970 Brasilia-DF; [e-mail]

ABSTRACT

This study was proposed to select minimum descriptors to the characterization of accessions of Capsicum spp. from Embrapa Hortaliças germplasm bank, which are characterized mainly by qualitative data. From this collection, 893 (420 C. annuum, 106 C. baccatum, 307 C. chinense and 60 C. frutescens) were considered because they were completely characterized by 56 morphological descriptors. The statistical treatment was performed by factorial multiple correspondence analysis, combined with estimation of matrices of genetic dissimilarity among accessions for multicategoric data. Lists of minimum descriptors were proposed for the whole collection and for the subcollections related to the four cultivated species. Around thirty minimum descriptors were selected for each subcollection, maintaining a minimum of 0.90 for the correlation between the dissimilarity matrices that consider all the descriptors or only the selected minimum descriptors. These proposed lists allow a reduction of around 50% in the number of initial descriptors. The filament colour, mature fruit colour, number of locules, fruit position, origin, pungency, fruit surface, stem length, plant height and fruit length were selected as minimum descriptors, both to the whole collection and their subcollections. Most of these descriptors are related to the fruit traits of sweet and chilli peppers, which are important for genetic breeding of Capsicum because are related to storage, processing, marketing and consumption of commercial derived products.

Keywords: qualitative data, multicategoric variables, multiple correspondence analysis, discarding variables.

RESUMO

Neste estudo buscou-se selecionar descritores mínimos para caracterização de acessos do banco de germoplasma de Capsicum spp. da Embrapa Hortaliças, caracterizado principalmente com base em dados de natureza qualitativa. Desta coleção foram considerados 893 acessos (420 C. annuum, 106 C. baccatum, 307 C. chinense e 60 C. frutescens), que estavam completamente caracterizados segundo 56 descritores morfológicos. O tratamento estatístico foi feito por análise fatorial de correspondência múltipla, combinada com estimação de matrizes de distâncias genéticas entre acessos, para dados multicategóricos. Foram propostas listas de descritores mínimos para a coleção completa e para as subcoleções relacionadas às quatro espécies. Cerca de trinta descritores foram selecionados para cada subcoleção, garantindo-se correlação mínima de 0,90 entre as matrizes de distâncias que consideram todos os descritores ou apenas os descritores mínimos selecionados. As listas propostas possibilitam reduções de quase 50% no número de descritores. Cor do filamento, cor do fruto maduro, número de lóculos, posição do fruto, procedência, pungência, superfície do fruto, comprimento da haste, comprimento da planta e comprimento do fruto foram eleitos como descritores mínimos, tanto para a coleção completa quanto para suas subcoleções. Grande parte destes descritores são características do fruto em pimentas e pimentões, importantes para o melhoramento genético de Capsicum, haja vista suas relações com armazenamento, processamento, comercialização e consumo dos produtos comerciais derivados.

Palavras-chave: dados qualitativos, variáveis multicategóricas, análise de correspondência múltipla, descarte de variáveis.

Plant genetic diversity is a valuable guarantee against possible adversaries that threaten the survival of biological species. Research on genetic resources and plant breeding is one of the innovation activities most relevant for agriculture sustainability. In this sense, the access, organization and maintenance of important collections of exotic variability and the significant efforts with collection, characterization, conservation and promotion for the use of native species of agricultural food interest have helped to create the bases of sustainability for various genetic breeding programs.

Germplasm characterization consists of programmed and systematic obtaining of data based on the assessment of characteristics that describe and differentiate the existing accessions (Almeida et al., 2005). This characterization therefore focuses basically on morphological and phenological aspects observed in the accessions and their comparison with lists of descriptive characteristics (descriptors). The state of the descriptors can vary among the accessions, beyond that foreseen on the lists, and some may be shown to be invariable for whole collections. It can sometimes be observed that certain descriptors have no value for identifying duplicates or reliable discrimination of the accessions.

There are several examples in the literature that show certain discriminatory incapacity of the traits used to classify accessions in germplasm banks. Onyilgha (1986) worked with the Dioscorea species and concluded that only 15 of 40 morphological traits were sufficient to discriminate the accessions. Bekele et al. (1994) selected only 23 descriptors from a list of 68 variables used to characterize a collection of cacao. Strapasson et al. (2000) discarded 86% of the descriptors used in characterizing of Paspalum sp. accessions.

Some descriptor lists, mostly prepared by research groups commissioned by Biodiversity International, the former IPGRI (International Plant Genetic Resources Institute), require taking a great number of observations, making it a very time-consuming and expensive task. Frankel & Brown (1984) criticized the extension and level of detailing on these lists. They point out that this effort did not bring the equivalent gains to the heavy workload imposed on curators and breeders. For these authors, the morphological and phenological data taking should be restricted to identifying duplicates and establishing nuclear collections.

Reliable methods and processes for germplasm characterization are fundamental to increase the use of the variability available. Its efficiency would anticipate the potential of each accession in the face of the desired genetic characteristics. For this, the use of multivariate techniques has been efficient not only in identifying the descriptors of greater interest, but also in discarding those that are not so relevant (Pereira, 1989). These procedures have been used in field crops such as Paspalum sp. (Strapasson et al., 2000), Capsicum chinense (Luz, 2007), tomato (Gonçalves et al., 2009) and Capsicum spp. (Ortiz et al., 2010; Sudré et al., 2010).

The objective of the present study was to select, from the descriptors available in the characterization of the Capsicum spp. collection of Embrapa Hortaliças, in Brazil, those essential for a reliable discrimination of the accessions, both for the whole collection and for the subcollections related to their four cultivated species (C. annuum, C. baccatum, C. chinense and C. frutescens). This will reduce efforts in characterizing the collections of this kind, based on smaller lists of descriptors.

MATERIAL AND METHODS

The data for this research were obtained at Embrapa Hortaliças that has a germplasm collection of chilli and sweet peppers (Capsicum spp.) with about 4,000 accessions. From this total, 893 accessions were considered in the study. They were brought from different regions of Brazil and abroad, and are completely characterized by 56 morphological descriptors (Table 1). As most of these descriptors are qualitative data, the other traits were also categorized, to gather them together in the same type of analysis. The 893 accessions are distributed in 420 C. annuum accessions, 106 C. baccatum accessions, 307 C. chinense accessions, and 60 C. frutescens accessions.

Multiple Correspondence Analysis (MCA) (Escofier & Pagès, 1992) was used to select the minimum descriptors. Analogically to the known Principle Components Analysis (PCA), applied to quantitative data, this analysis also allows to study a population of n individuals assessed according to p qualitative variables. The selection of descriptors was based mainly on the approach proposed by Curry (1993) and Dias (1994), based on method of Jolliffe (1973), substituting PCA for MCA. According to this proposal, after carrying out the analysis with all the descriptors (complete analysis), the one that most contributed to the last factorial axis is discarded. Given that importance of the principal components or factorial axes decreases from the first to the last, this one explains usually a very small portion of the total variance. For this, the descriptor with the highest coefficient on the principal axis with the smallest eigenvalue (the last axis) can be discarded (Pereira, 1989). For each discarding, another analysis is carried out with the remaining descriptors, and thus successively until the ranking of the descriptors is established by their importance. Fifty-five analyses (p-1) were carried out using the corresp procedure from the SAS computation system (SAS Institute, 2009).

Based on the results of these analyses, three selection criteria for minimum descriptors were assessed: C1 – selection according the proposal by Jollife (1973) method, based on the inverse order of the descriptor for the p^th (last) factorial axis (O'_p); C2 - selection based on the mean of the contribution orders of the descriptor (O_s) on the first three factorial axes in the complete analysis, and O'_p, so that O_s = (O₁ + O₂ + O₃ + O'_p)/4; and C3 - selection based on the weighted mean of the contribution orders of the descriptor (O_z) for the three first axes, with weights (w₁, w₂ and w₃) defined by the respective eigenvalues in the complete analysis: O_z = (w₁O₁ + w₂O₂ + w₃O₃)/(w₁+w₂+w₃).

Before applying and assessing the three criteria, uniform size was established for the list of minimum descriptors to be constituted within each group (complete collection or subcollections). This size (around thirty descriptors) was defined having as target a correlation equal or greater to 0.90 between the dissimilarity matrices of accessions estimated with all the descriptors (M) or only with the minimum descriptors (E). The dissimilarity index proposed by Cruz & Carneiro (2006) was used for multi-categorical variables: [d_ii' = D/(C+D)], where i and i' correspond to a pair of accessions (i e i' = 1, 2, ..., n), C is the number of category agreements, and D is the number of disagreements. This was performed in the Genes computer program (Cruz, 2006).

The correlation between M and E dissimilarity matrices, for each germplasm group, taking account the different criteria, was used as the measure of the quality of the minimum descriptor lists. The best criterion, chosen to establish the definitive list of minimum descriptors, was that which resulted in the highest correlation between these matrices. The significance of these correlations was assessed by the Mantell test, carried out on the NTSys software, with 10,000 randomizations (Rohlf, 2005).

RESULTS AND DISCUSSION

Capsicum spp. - Table 2 shows the ranking of the descriptors for the whole germplasm collection, according to their contributions to the last factorial axis in the successive analyses (MCA). It was observed in the analysis with all the descriptors that the variable "species" (Esp) was that most contributed to explain this factor axis and, in principle, it would be of less importance for the characterization of these accessions. In theory, it means that the information associated to this variable is already contemplated by other variables that determine the previous axes. And in this case, this certainly happened because the identification of the species is related mainly to differences in flower characteristics (Moscone et al., 2007; Ince et al., 2009).

On the other hand, this variable (Esp) showed the second highest contribution to the first factorial axis, and the greatest contribution to the second and third axes. Partial results such as these are not presented here, but are available in Silva (2008), and also for the four subcollections. This same behavior was observed in other variables: fruit weight (PFt), with the greatest contribution to the first factorial axis; corolla colour (CrC) - fourth greatest contribution to the first axis; calyx annular constriction (CCl) - second greatest contribution to the second axis; and corolla spot colour (CCr) - second greatest contribution to the third axis. Because of this it was decided to also consider the participation of the descriptors in the first three factorial axes (criteria C2 and C3), instead of discarding variables only with base on the C1 criterion (Jollife, 1973).

Assessment of the contributions of the descriptors to the first three factorial axes in the complete MCA showed that about 50% of the descriptors with the greatest contributions explained 90% or more of the total variability associated to each axis (results available in Silva, 2008). This is only possible because of the considerable amount of redundancy or association among the variables; that is, some descriptors capture information of the same nature about the variability of the accessions. However, for the purpose of the study, this redundancy was the guarantee that the number of descriptors needed to characterize the germplasm could be reduced without great harm to the representation of their genetic variability. In this case, it was observed that the accumulated contribution of the first variable (PFt) up to the 28^th (PFr), on the first factorial axis, was 92.5% (Silva, 2008). Therefore to guarantee minimum explanation of 90% of the variability captured by the first and main spectre of the factorial analysis, about thirty variables should be selected as minimum descriptors.

The ranking of the descriptors according to the three selection criteria (C1, C2 and C3) for this collection is shown in Table 3. The best descriptive ability is associated to the descriptor with the lowest of these order. Taking account the thirty best classified descriptors in each criterion, significant correlations (p<0.01) among the matrices M (estimated with all the descriptors) and E (estimated only with the minimum descriptors) were observed, with magnitude of 0.80 (C1), 0.89 (C2) and 0.87 (C3). Therefore, the criterion C2 was chosen to select the minimum descriptors.

The definitive list of the minimum descriptors for Capsicum spp. was defined as: species (Esp), fruit wall thickness (EFt), fruit width (LFt), number of flowers per axil (NFA), fruit weight (PFt), fruit position (PsF), origin (Pro), ripe fruit persistence (PtF), calyx margin (MgC), fruit shape (FFt), fruit shape at pedicel attachment (OFt), number of seeds per fruit (NSF), anther colour (CAt), number of locules (NLl), filament colour (CFn), placenta length (CPl), corolla colour (CrC), mature fruit colour (CFM), fruit shape at blossom end (FPF), leaf shape (FmF), pungency (Pug), fruit pedicel length (CpP), plant canopy width (LPt), stigma exsertion (PEt), fruit cross-sectional corrugation (STv), aroma (Arm), days to fruiting (DFt), fruit surface (SFt), nodal anthocyanin (AtN), plant growth habit (Hbt) and flower position (PFr).

Regarding the presence of the variable "species" (Esp) among the minimal descriptors of this collection, it is needed to point out, firstly, that if the accession is a cultivated material or if the species has already been identified, its characterization can be realized directly with base on the list of minimum descriptors of its species. Otherwise, the inclusion of this variable should be interpreted as the need to maintain, among the minimum descriptiors of a generic collection, all the morphological attributes involved in the identification key of species in Capsicum spp. This naturally is irrelevant within the specific subcollections.

Even though the variable "origin" (Pro) is not an intrinsic attribute to the morphology of the accessions, although it is important passport data, its inclusion in this list shows that, regardless of the species, the geographic origin of the germplasm is directly associated to their genetic divergence. As will be shown further, this fact was also observed in the study of the subcollections. According to Galwey (1995), good passport data (e.g. geographic origin) are probably unrivaled in supplying concise and inexpensive information about the structure of the genetic variability of a germplasm collection.

The analysis of this collection further highlights the high proportion of minimum descriptors related to the fruit characteristics. This is interesting because they are traits commonly evaluated in the chilli and sweet pepper consumer markets (Onoyama et al., 2010). Thus, the characterization of accessions by these descriptors is directly linked to the interests of the Capsicum breeders (e.g. identification of potential parents for their crossing programs). Furthermore, this corroborates the results obtained by Teixeira (1996) and Sudré (2003), who emphasized length, diameter and mean weight of the fruit as the most important descriptors for genetic divergence in Capsicum.

Capsicum annuum -Considering only the C. annuum accessions, the variable with greatest contribution to the last factorial axis in the analysis with all the descriptors was the corolla colour (CrC). Thus, according criterium C1 (Jollife, 1973), this variable could be discarded. However, as already discussed, this descriptor also showed the fourth greatest contribution to the second factorial axis, and the greatest to the third axis. This fact reiterates the need to aggregate information on the contribution of the descriptors in the first axes of the MCA.

Some descriptors did not contribute to the factorial axes, that is, they were invariant in this germplasm group: calyx annular constriction (CCl), male-sterility (McE), neck at base of fruit (PcF) and, obviously, species (Esp). Thus they were discarded. In a similar case, Simpson et al. (2002) highlighted the trait leaf colour when separating cultivated peanut accessions, because the descriptor was invariable in 681 accessions. However, it should be emphasized that in this type of study commonly the relative contribution of each factorial axis is fairly small, bearing in mind the great number of variables and categories per variable.

The variables that most contributed to the first factorial axis of the analysis with all the descriptors in this group of accessions were: fruit weight (PFt), fruit width (LFt), fruit wall thickness (EFt), fruit shape at pedicel attachment (OFt), pungency (Pug), placenta length (CPl), fruit shape (FFt), number of locules (NLl), branching habit (DsR), and fruit shape at blossom end (FPF). It was again observed that most of these were related to fruit characteristics, which was an interesting fact because these traits are used in breeding of the species, especially in the sweet pepper group. This may have resulted further from the collect and domestication processes that overvalue this type of characteristic. The same tendency was observed in the second factorial axis, with greatest contribution from fruit weight (PFt), fruit length (CFt), fruit width (LFt), corolla colour (CrC), flower position (PFr), fruit position (PsF), and fruit wall thickness (EFt). In the Capsicum genus, fruit shape, and immature and mature fruit colour are important traits for the consumer market (Onoyama et al., 2010; Sudré et al., 2006).

It was also observed in this group (C. annuum) that 50% of the descriptors explained almost 90% of the variability captured by the first factorial axis (Table 4); that is, the accumulated contribution of the first variable up to the 28^th variable corresponded to approximately 90%. Thus selecting at least the first 29 descriptors (about thirty variables), guarantees again this minimal explanation of the variability associated to the first principal axis of the analysis.

There was significant correlation (p<0.01) among M and E dissimilarity matrices for all criteria of descriptor selection (0.84 for C1 criterion, and 0.91 for C2 and C3). The C2 criterion was chosen because of its better performance also in the whole collection study. The list of minimum descriptors for this group was defined as: fruit length (CFt), calyx pigmentation (Pgl), fruit wall thickness (EFt), fruit width (LFt), origin (Pro), fruit weight (PFt), stem colour (CrH), corolla colour (CrC), immature fruit colour (CFI), plant growth habit (Hbt), filament colour (CFn), stem length (CHt), anther colour (CAt), fruit shape at pedicel attachment (OFt), fruit shape at blossom end (FPF), tillering (BtF), fruit position (PsF), leaf density (DsF), leaf colour (CrF), fruit shape (FFt), number of seeds per fruit (NSF), pungency (Pug), fruit surface (SFt), calyx margin (MgC), mature fruit colour (CFM), plant height (CPt), flower position (PFr), number of locules (NLl), stem pubescence (PHt), and plant canopy width (LPt). In this case, eleven of these are also fruit characteristics. Immature fruit colour, fruit shape, and length and width of the fruit are related mainly to fresh fruit commercialization; wall thickness and number of seeds per fruit are important traits when selling the dried product; and mature fruit colour is interesting in both forms of commercialization.

Capsicum baccatum - Observation of the C. baccatum accessions showed that some descriptors had no descriptive importance (invariant) for the characterization of the species: male sterility (McE), calyx annular constriction (CCl), seed colour (CSt), seed surface (SSt), corolla spot colour (CCr) and, obviously, species (Esp). Thus these variables should not be part of the respective descriptor list. In this group, the variables that most contributed to the first factorial axis in the complete analysis were: fruit weight (PFt), fruit length (CFt), ripe fruit persistence (PtF), fruit wall thickness (EFt), number of seeds per fruit (NSF), fruit width (LFt), days to fruiting (DFt), leaf pubescence (PbF), nodal anthocyanin (AtN), and number of flowers per axil (NFA). Also in this case, fruit-related descriptors presented greater variability that is a marked characteristic in the Capsicum genus.

From the ranking of the descriptors according to the three criteria (Table 5) and the respective lists of selected descriptors, the criterion which generated the highest correlation (0.94) between M and E matrices was again C2. The C1 and C3 criteria produced the correlations 0.77 and 0.64, respectively. All these correlations were significant (p<0.01). Thus, the minimum descriptor list was defined by C2 criterion as: fruit shape at blossom end (FPF), fruit position (PsF), fruit length (CFt), fruit weight (PFt), fruit shape (FFt), fruit wall thickness (EFt), fruit surface (SFt), corolla colour (CrC), stem length (CHt), plant height (CPt), number of locules (NLl), immature fruit colour (CFI), origin (Pro), fruit pedicel length (CpP), pungency (Pug), days to flowering (DFl), number of flowers per axil (NFA), tillering (BtF), filament colour (CFn), number of seeds per fruit (NSF), fruit width (LFt), leaf pubescence (PbF), stigma exsertion (PEt), stem pubescence (PHt), fruit shape at pedicel attachment (OFt), placenta length (CPl), mature fruit colour (CFM), days to fruiting (DFt), branching habit (DsR), and plant growth habit (Hbt).

As observed in the previous analyses, twelve of the thirty descriptors selected are fruit characteristics. This corroborates the importance of these traits to discriminate accessions in germplasm bank of Capsicum. Thus maintaining these descriptors not only guarantees parsimony in the germplasm characterization but also supplies information of immediate interest to the breeders, because it is linked to the market demands regarding the cultivars and commercial products derived from these peppers (Sudré et al., 2006).

Capsicum chinense - Table 6 shows the ranking of the variables according to the criteria for the composition of the minimal descriptor list for C. chinense. Choosing thirty variables according to each criterion, the correlations between M and E matrices, all significant (p<0.01) were 0.82 for C1, 0.88 for C2, and 0.88 for C3. Thus, once again, the C2 and C3 criteria were outstanding. Due to the same reasons previously reported, the following descriptors were chosen to compose the list according C2 criterion: flower position (PFr), number of locules (NLl), fruit pedicel length (CpP), mature fruit colour (CFM), fruit surface (SFt), anther colour (CAt), fruit position (PsF), fruit length (CFt), fruit wall thickness (EFt), fruit shape at pedicel attachment (OFt), fruit shape (FFt), origin (Pro), fruit weight (PFt), number of seeds per fruit (NSF), stem shape (FHt), days to fruiting (DFt), fruit width (LFt), ripe fruit persistence (PtF), days to flowering (DFl), calyx margin (MgC), leaf pubescence (PbF), plant canopy width (LPt), fruit shape at blossom end (FPF), stem diameter (DHt), plant height (CPt), leaf shape (FmF), immature fruit colour (CFI), stem length (CHt), filament colour (CFn), placenta length (CPl), corolla shape (FmC), and pungency (Pug). Again, of the 32 descriptors selected twelve are fruit traits, corroborating the previous results from C. annuum and C. baccatum, and those reported by Teixeira (1996) and Sudré (2003).

Capsicum frutescens - Table 7 shows the rankings of the variables in this germplasm group, according to the criteria for establishment of the minimum descriptors. The correlations between M and E matrices, all significant (p<0.01), were: 0.83 for C1, 0.89 for C2, and 0.92 for C3. Thus, in this case the advantage was in favor of C3 criterion, that resulted in the following list of minimal descriptors: mature fruit colour (CFM), nodal anthocyanin (AtN), fruit position (PsF), origin (Pro), days to flowering (DFl), ripe fruit persistence (PtF), aroma (Arm), number of flowers per axil (NFA), stem shape (FHt), fruit cross-sectional corrugation (STv), plant height (CPt), branching habit (DsR), plant canopy width (LPt), days to fruiting (DFt), anther colour (CAt), stigma exsertion (PEt), fruit surface (SFt), stem length (CHt), stem colour (CrH), filament colour (CFn), number of locules (NLl), leaf colour (CrF), tillering (BtF), calyx margin (MgC), leaf density (DsF), corolla shape (FmC), pungency (Pug), fruit length (CFt), flower position (PFr), and stem diameter (DHt). Again, about a third of these descriptors are fruit traits, reinforcing the results reported previously.

Synthesis of the minimum descriptors - Table 8 shows the minimal descriptor lists proposed generically for Capsicum spp., and, in particular, for the cultivated species (C. annuum, C. baccatum, C. chinense and C. frutescens). Stem length (CHt), plant height (CPt) and fruit length (CFt) were added to the listing for Capsicum spp. because these descriptors were always included in the listings established for the specific subcollections.

Eight of the 56 descriptors studied (fruit blossom end appendage, calyx annular constriction, corolla spot colour, seed colour, male-sterility, varietal mixture condition, neck at base of fruit, and seed surface) were not essential for characterizing the accessions of this germplasm. Some of these, however, can be occasionally considered in this type of characterization; for example, cayx annular constriction that is one of the main traits for morphological distinction between C. chinense and C. frutescens (Sudré et al., 2006). Five other descriptors were selected exclusively for one of the collections ("species" for Capsicum spp.; calyx pigmentation for C. annuum; and nodal anthocyanin, aroma and fruit cross sectional corrugation for C. frutescens), although some with relatively little importance (e.g. Pgl and STv).

These facts showed the possibility of reducing the list of descriptors currently adopted at the Capsicum germplasm bank of Embrapa Hortaliças, and without harming the representation of their genetic variability. In general, the minimal descriptor lists proposed in the present study (Table 8) allow approximately 50% of the original descriptors to be discarded.

It is pointed out that ten of the minimal descriptors indicated (Table 8) are present in the five established lists: filament colour (CFn), mature fruit colour (CFM), number of locules (NLl), fruit position (PsF), origin (Pro), pungency (Pug), fruit surface (SFt), stem length (CHt), plant height (CPt), and fruit length (CFt). These descriptors were shown therefore to be relevant for germplasm characterization in the four Capsicum species studied.

A further fourteen descriptors also deserve mention because were included in the lists of three of the four species: tillering (BtF), anther colour (CAt), immature fruit colour (CFI), days to flowering (DFl), days to fruiting (DFt), fruit wall thickness (EFt), fruit shape (FFt), fruit shape at blossom end (FPF), plant canopy width (LPt), calyx margin (MgC), number of seeds per fruit (NSF), fruit shape at pedicel attachment (OFt), flower position (PFr), and fruit weight (PFt). Only three of these were not repeated in the generic Capsicum spp. list (BtF, CFI and DFl). Another fifteen were part of two of these lists, and eight (CPl, CpP, CrC, FmF, Hbt, NFA, PtF and PEt) are also on the generic list. Thus, the minimal descriptors selected for the whole collection are in general represented among those of the subcollections. In spite of this, according to Valls (2007), the most suitable descriptors to discriminate among species of a genus tend to be the least suitable for differentiating cultivars within of species.

Another fact that deserves further emphasis is that most of the descriptors selected are fruit characteristics. Given the importance of these traits for Capsicum breeding, it was found that it is possible to characterize these germplasm collections to meet simultaneously the criteria of representing genetic variability and the most immediate interests of the breeder. In this sense, if the germplasm characterization takes into account traits of interest for breeding, the possibilities of its use are increased, as stipulated in public policies on conservation and use of plant genetic sources (FAO, 1991).

From the methodological point of view, the results of this study show, firstly, that the multiple correspondence factorial analysis (MCA) is an useful alternative in the multivariate analysis of data from germplasm collections, especially when most of the descriptors have a qualitative nature. Secondly, it enabled to state that the criteria of discarding of variables based on the original proposal by Jolliffe (1973) is less efficient than the alternatives that value more the contribution of the variables on the first factorial axes.

ACKNOWLEDGEMENTS

The authors thank Embrapa Hortaliças by providing the data, and National Council for Scientific and Technological Development (CNPq) for offering the first author the conditions to conclude this study.

REFERENCES

ALMEIDA CMCV; DIAS LAS; OKABE ET; MEDEIROS JRP. 2005. Variability in genetic resources of cacao in Rondônia, Brazil. Crop Breeding and Applied Biotechnology 5: 318-324.

BEKELE FL; KENNEDY AJ; MCDAVID C; LAUCKNER FB; BEKELE I. 1994. Numerical taxonomic studies on cacao (Theobroma cacao L.) in Trinidad. Euphytica 75: 231-240.

CRUZ CD. 2006. Programa Genes. Software. Versão Windows: aplicativo computacional em genética e estatística. Viçosa: UFV. (CD-ROM).

CRUZ CD; CARNEIRO PCS. 2006. Modelos biométricos aplicados ao melhoramento genético. 2 ed. Viçosa: UFV. v. 2, 585p.

CURY R. 1993. Dinâmica evolutiva e caracterização de germoplasma de mandioca (Manihot esculenta Crantz) na agricultura autóctone do sul do Estado de São Paulo. Piracicaba: USP-ESALQ. 103p (Dissertação mestrado).

DIAS LAS. 1994. Divergência genética e fenética multivariada na predição de híbrido e preservação de germoplasma de cacau (Theobroma cacao L.). Piracicaba: USP-ESALQ. 94p. (Tese doutorado).

ESCOFIER B; PAGÈS J. 1992. Análisis factoriales simples y múltiples: objetivos, métodos e interpretación. Bilbao: Servicio Editorial de la Universidad del País Vasco. 285p.

FAO. 1991. Relatório da quarta reunião da Comissão de Recursos Fitogenéticos. Roma: FAO. 96p.

FRANKEL OH; BROWN AHD. 1984. Plant genetic resources today: a critical reappraisal. In: HOLDEN JHW; WILLIAMS JT (eds). Crop genetic resources: conservation and evaluation. London: George Allen & Unwin, p. 249-257.

GALWEY NW. 1995. Verifying and validating the representativeness of a core collection. In: HODGKIN T; BROWN AHD; VAN HINTUM TJL; MORALES EAV (eds). Core collections of plant genetic resources. New York: J. Wiley. p. 187-198.

GONÇALVES LS; RODRIGUES R; AMARAL JUNIOR AT; KARASAWA M; SUDRÉ CP. 2009. Heirloom tomato gene bank: assessing genetic divergence based on morphological, agronomic and molecular data using a Ward-modified location model. Genetics and Molecular Research, 8: 364-374.

INCE AG; KARACA M; ONUS AN. 2009. Development and utilization of diagnostic DAMD-PCR markers for Capsicum accessions. Genetic Resources and Crop Evolution 56: 211-221.

JOLLIFFE IT. 1973. Discarding variables in a principal component analysis, II Real data. Applied Statistics 22: 21-31.

LUZ FJF. 2007. Caracterizações morfológica e molecular de pimentas (Capsicum chinense Jack.). Jaboticabal: UNESP-FCAV. 150p (Tese doutorado).

MOSCONE EA; SCALDAFERRO MA; GRABIELE M; CECCHINI NM; SÁNCHEZ GARCÍA Y; JARRET R; DAVIÑA JR; DUCASSE DA; BARBOZA GE; EHRENDORFER F. 2007. The evolution of Chili Peppers (Capsicum - Solanaceae): a cytogenetic perspective. VI International Solanaceae Conference: Genomics Meets Biodiversity. Acta Horticulturae 745: 137-170.

ONOYAMA SS; REIFSCHNEIDER FJB; MOITA AW; SOUZA GS. 2010. Atributos de hortaliças sob a ótica de consumidores: estudo de caso do pimentão no Distrito Federal. Horticultura Brasileira 28: 124-132.

ONYILAGHA JC. 1986. Numerical analysis of variation among Nigerian Dioscorea rotunda accessions. Euphytica 35: 413-419.

ORTIZ R; DE LA FLOR FD; ALVARADO G; CROSSA J. 2010. Classifying vegetable genetic resources. A case study with domesticated Capsicum spp. Scientia Horticulturae 126: 186-191.

PEREIRA AV. 1989. Utilização de análise multivariada na caracterização de germoplasma de mandioca (Manihot esculenta Crantz). Piracicaba: USP-ESALQ. 180p (Tese doutorado).

ROHLF FJ. 2005. NTSYS-PC: Numerical taxonomy and multivariate analysis system. Version 2.2: software. New York: Exeter/Setauket.

SAS INSTITUTE. 2009. SAS^® System for Windows. Version 9.0: software. SAS Institute: Cary, NC.

SILVA WCJ. 2008. Escolha de descritores mínimos e estabelecimento de coleções nucleares em Capsicum spp. Goiânia: UFG. 139p (Tese doutorado).

SIMPSON CE; HOWARD ER; HIGGINS DL; THOMAS GD. 2002. Catalog of minimum descriptors of Arachis hypogaea L. germplasm collections from South America. College Station: Texas A&M University, p. 1977-1982.

STRAPASSON E; VENCOVSKY R; BATISTA LAR. 2000. Seleção de descritores na caracterização de germoplasma de Paspalum sp. por meio de componentes principais. Revista Brasileira de Zootecnia 29: 373-381.

SUDRÉ CP. 2003. Divergência genética e avaliação da resistência à mancha bacteriana em Capsicum spp. Campos dos Goytacazes: UENF. 112p (Tese doutorado).

SUDRÉ CP; CRUZ CD; RODRIGUES R; RIVAL EM; AMARAL JÚNIOR AT; SILVA DJH; PEREIRA TNS. 2006. Variáveis multicategóricas na determinação da divergência genética entre acessos de pimenta e pimentão. Horticultura Brasileira 24: 88-93.

SUDRÉ CP; GONÇALVES LSA; RODRIGUES R; AMARAL JÚNIOR AT; RIVA-SOUZA EM; BENTO CS. 2010. Genetic variability in domesticated Capsicum spp. as assessed by morphological and agronomic data in mixed staatistical analysis. Genetics and Molecular Research 9: 283-294.

TEIXEIRA R. 1996. Diversidade em Capsicum: análise molecular, morfoagronômica e química. Viçosa: UFV. 81p (Dissertação mestrado).

VALLS JFM. 2007. Caracterização de recursos genéticos vegetais. In: Nass LL (ed). Recursos genéticos vegetais. Brasília: Embrapa Recursos Genéticos e Biotecnologia, v. 1, cap. 8, p. 281-305.

(Received on April 16, 2012; accepted on February 18, 2013)

References

• ALMEIDA CMCV; DIAS LAS; OKABE ET; MEDEIROS JRP. 2005. Variability in genetic resources of cacao in Rondônia, Brazil. Crop Breeding and Applied Biotechnology 5: 318-324.
• BEKELE FL; KENNEDY AJ; MCDAVID C; LAUCKNER FB; BEKELE I. 1994. Numerical taxonomic studies on cacao (Theobroma cacao L.) in Trinidad. Euphytica 75: 231-240.
• CRUZ CD. 2006. Programa Genes Software. Versão Windows: aplicativo computacional em genética e estatística. Viçosa: UFV. (CD-ROM).
• CRUZ CD; CARNEIRO PCS. 2006. Modelos biométricos aplicados ao melhoramento genético 2 ed. Viçosa: UFV. v. 2, 585p.
• CURY R. 1993. Dinâmica evolutiva e caracterização de germoplasma de mandioca (Manihot esculenta Crantz) na agricultura autóctone do sul do Estado de São Paulo Piracicaba: USP-ESALQ. 103p (Dissertação mestrado).
• DIAS LAS. 1994. Divergência genética e fenética multivariada na predição de híbrido e preservação de germoplasma de cacau (Theobroma cacao L.) Piracicaba: USP-ESALQ. 94p. (Tese doutorado).
• ESCOFIER B; PAGÈS J. 1992. Análisis factoriales simples y múltiples: objetivos, métodos e interpretación. Bilbao: Servicio Editorial de la Universidad del País Vasco. 285p.
• FAO. 1991. Relatório da quarta reunião da Comissão de Recursos Fitogenéticos Roma: FAO. 96p.
• FRANKEL OH; BROWN AHD. 1984. Plant genetic resources today: a critical reappraisal. In: HOLDEN JHW; WILLIAMS JT (eds). Crop genetic resources: conservation and evaluation London: George Allen & Unwin, p. 249-257.
• GALWEY NW. 1995. Verifying and validating the representativeness of a core collection. In: HODGKIN T; BROWN AHD; VAN HINTUM TJL; MORALES EAV (eds). Core collections of plant genetic resources New York: J. Wiley. p. 187-198.
• GONÇALVES LS; RODRIGUES R; AMARAL JUNIOR AT; KARASAWA M; SUDRÉ CP. 2009. Heirloom tomato gene bank: assessing genetic divergence based on morphological, agronomic and molecular data using a Ward-modified location model. Genetics and Molecular Research, 8: 364-374.
• INCE AG; KARACA M; ONUS AN. 2009. Development and utilization of diagnostic DAMD-PCR markers for Capsicum accessions. Genetic Resources and Crop Evolution 56: 211-221.
• JOLLIFFE IT. 1973. Discarding variables in a principal component analysis, II Real data. Applied Statistics 22: 21-31.
• LUZ FJF. 2007. Caracterizações morfológica e molecular de pimentas (Capsicum chinense Jack.) Jaboticabal: UNESP-FCAV. 150p (Tese doutorado).
• MOSCONE EA; SCALDAFERRO MA; GRABIELE M; CECCHINI NM; SÁNCHEZ GARCÍA Y; JARRET R; DAVIÑA JR; DUCASSE DA; BARBOZA GE; EHRENDORFER F. 2007. The evolution of Chili Peppers (Capsicum - Solanaceae): a cytogenetic perspective. VI International Solanaceae Conference: Genomics Meets Biodiversity. Acta Horticulturae 745: 137-170.
• ONOYAMA SS; REIFSCHNEIDER FJB; MOITA AW; SOUZA GS. 2010. Atributos de hortaliças sob a ótica de consumidores: estudo de caso do pimentão no Distrito Federal. Horticultura Brasileira 28: 124-132.
• ONYILAGHA JC. 1986. Numerical analysis of variation among Nigerian Dioscorea rotunda accessions. Euphytica 35: 413-419.
• ORTIZ R; DE LA FLOR FD; ALVARADO G; CROSSA J. 2010. Classifying vegetable genetic resources. A case study with domesticated Capsicum spp. Scientia Horticulturae 126: 186-191.
• PEREIRA AV. 1989. Utilização de análise multivariada na caracterização de germoplasma de mandioca (Manihot esculenta Crantz). Piracicaba: USP-ESALQ. 180p (Tese doutorado).
• ROHLF FJ. 2005. NTSYS-PC: Numerical taxonomy and multivariate analysis system Version 2.2: software. New York: Exeter/Setauket.
• SAS INSTITUTE. 2009. SAS^® System for Windows Version 9.0: software. SAS Institute: Cary, NC.
• SILVA WCJ. 2008. Escolha de descritores mínimos e estabelecimento de coleções nucleares em Capsicum spp. Goiânia: UFG. 139p (Tese doutorado).
• SIMPSON CE; HOWARD ER; HIGGINS DL; THOMAS GD. 2002. Catalog of minimum descriptors of Arachis hypogaea L. germplasm collections from South America College Station: Texas A&M University, p. 1977-1982.
• STRAPASSON E; VENCOVSKY R; BATISTA LAR. 2000. Seleção de descritores na caracterização de germoplasma de Paspalum sp. por meio de componentes principais. Revista Brasileira de Zootecnia 29: 373-381.
• SUDRÉ CP. 2003. Divergência genética e avaliação da resistência à mancha bacteriana em Capsicum spp. Campos dos Goytacazes: UENF. 112p (Tese doutorado).
• SUDRÉ CP; CRUZ CD; RODRIGUES R; RIVAL EM; AMARAL JÚNIOR AT; SILVA DJH; PEREIRA TNS. 2006. Variáveis multicategóricas na determinação da divergência genética entre acessos de pimenta e pimentão. Horticultura Brasileira 24: 88-93.
• SUDRÉ CP; GONÇALVES LSA; RODRIGUES R; AMARAL JÚNIOR AT; RIVA-SOUZA EM; BENTO CS. 2010. Genetic variability in domesticated Capsicum spp. as assessed by morphological and agronomic data in mixed staatistical analysis.  Genetics and Molecular Research 9: 283-294.
• TEIXEIRA R. 1996. Diversidade em Capsicum: análise molecular, morfoagronômica e química Viçosa: UFV. 81p (Dissertação mestrado).
• VALLS JFM. 2007. Caracterização de recursos genéticos vegetais. In: Nass LL (ed). Recursos genéticos vegetais Brasília: Embrapa Recursos Genéticos e Biotecnologia, v. 1, cap. 8, p. 281-305.

Publication Dates

• Publication in this collection
  18 July 2013
• Date of issue
  June 2013

History

• Received
  16 Apr 2012
• Accepted
  18 Feb 2013