Average Linkage Algorithm Research Articles

A novel unsupervised ensemble framework using concept-based linguistic methods and machine learning for twitter sentiment analysis

• A Novel ensemble/cooperative framework based on concept-based and clustering is proposed to perform Twitter sentiment Analysis. • It employs majority voting, tie breaker criteria, and linguistic rules in concept-based module. • Comparative analysis between clustering and classification is presented when integrated with concept based methods. • It presents the performance of feature representation methods (Boolean and TF-IDF). • Experimental results on Twitter Datasets revealed better performance of proposed framework. Concept-based sentiment analysis (CBSA) methods have gained prominence in natural language processing in recent years. These methods consider the underlying semantic meanings of text to perform different tasks such as Twitter sentiment analysis (assigning positive, negative, or neutral sentiment to Tweets). CBSA is superior to traditional statistical methods for accurately discovering sentiment labels. Due to a limited knowledge base, these methods are unable to identify the sentiment polarity of all kinds of text. Therefore, supervised learning techniques are mostly ensembled with CBSA methods to classify the whole text. These techniques require labeled data. It is a tedious and time-consuming task due to the manually labeling of large datasets (Such as Twitter datasets). Therefore, an unsupervised learning mechanism can be a better alternative to solve this problem. In this paper, a novel unsupervised learning framework based on Concept-based and hierarchical clustering is proposed for Twitter sentiment analysis. Popular hierarchical clustering methods including single linkage, complete linkage, and average linkage algorithms are ensembled serially. Two different feature representation methods including Boolean and Term frequency-inverse document frequency (TF-IDF) are investigated. We have also experimented with Well-known classifiers (Naive Bayes, Neural Network) for a fair comparison. Accuracy measure (proportion of correct predictions) is used to evaluate the performance of understudied techniques. It is empirically shown that the performance of unsupervised learning techniques is comparable with supervised learning techniques.

Coal elemental (compositional) data analysis with hierarchical clustering algorithms

The modes of occurrence for elements in coal are extremely important for deciphering geological process of coal formation and for anticipating the technological behavior and environmental and health impacts derived from coal utilization. Hierarchical clustering algorithm has been widely adopted to investigate the modes of occurrence of elements in coal. The traditional statistics (e.g., Pearson correlation, Euclidean distance) for the elemental data of coal may lead to misinterpretation because the elemental data of coal are of compositional nature and follow the rules of Aitchison geometry. This work applied log-ratio transformations in order to overcome this problem. Different hierarchical clustering algorithms with various data transformations can infer modes of occurrence for coal elements, but which algorithm is optimum deserves to be investigated. In this paper, we discuss four commonly used hierarchical clustering algorithms utilizing pivot coordinates and weighted symmetric pivot coordinates (WSPC), two types of log-ratio transformations, to infer modes of occurrence of elements in coal, based on published coal elemental data. Results showed that the Pearson correlation produces more meaningful results than the Euclidean distance in clustering rare earth elements and Y. WSPC produces more interpretable results than those from pivot coordinates transformed data for these coal elemental data. Compared with the single, complete, and centroid, the average-linkage algorithm is indeed the optimum.

Pengelompokkan Tingkat Kriminalitas di Indonesia Menggunakan Algoritma Average Linkage

Crime needs to be analyzed and grouped so that the act does not cause harm either ecologically or psychologically. The statistical method that can be used to classify crime is the Average Linkage Algorithm. The study aims to group and analyze the characteristics of criminal cases in Indonesia. From the results of the analysis, 3 clusters were formed based on the average of each cluster. Cluster 1 consists of Aceh, West Sumatra, Riau, Jambi, South Sumatra, Bengkulu, Lampung, Kep. Bangka Belitung, Kep. Riau, West Java, Central Java, DI Yogyakarta, East Java, Banten, Bali, West Nusa Tenggara, East Nusa Tenggara, West Kalimantan, Central Kalimantan, South Kalimantan, East Kalimantan, North Sulawesi, Central Sulawesi, South Sulawesi, Southeast Sulawesi, Gorontalo, Maluku, North Maluku and Papua. Cluster 2 consists of North Sumatra while Cluster 3 consists of Metro Jaya. The grouping results are the basis of the government, apparatus, and the community in implementing the handling of criminal acts that occur in each cluster area so that prevention can minimize the losses caused by these crimes.

Short Communication: Leaf architectural analysis of confusing Syzygium species: Syzygium aqueum (Burm.f.) Alston and Syzygium samarangense (Blume) Merr. & L.M.Perry (Myrtaceae)

Abstract. Viacrucis III JDL, Buot Jr. IE. 2021. Short Communication: Leaf architectural analysis of confusing Syzygium species: Syzygium aqueum (Burm.f.) Alston and Syzygium samarangense (Blume) Merr. & L.M.Perry (Myrtaceae). Biodiversitas 22: 3341-3348. Among species of Myrtaceae family which has economic and ecological importance are the two Syzygium species, Syzygium aqueum and Syzygium samarangense. Both species are cultivated throughout the tropics because of their edible fruits and medicinal properties of various plant parts. However, despite their wide utilization, the two species are often confusing due to their morphological similarity. In this study, leaf architecture of S. aqueum and S. samarangense have been evaluated to delineate the two confusing taxa. Selected laminal and venation characters were compared. Results showed many similar characters shared between the two species, however, laminar size, apex and base shape, areolation, and freely ending veinlets (FEVs) were found to be distinct. Multivariate analyses, cluster analysis employing Gower similarity index, using unweighted pair-group average (UPGMA) and single linkage (nearest neighbor) algorithm, and ordination using Principal Coordinates Analysis (PCoA), illustrated that they are grouped separately. The variations in laminal and venation characters are additional set of information that will allow accurate differentiation of the two species. However, it is recommended that additional taxonomical analysis using various methods, such as molecular techniques, be employed and characters, such as floral morphology, be studied to contribute in the resolution of the taxonomic complexity of this group and other taxonomic controversies

Efficient text feature extraction by integrating the average linkage and K-medoids clustering

By clustering feature words, we can not only simplify the dimension of feature subsets, but also eliminate the redundancy of the feature. However, for a feature set with very large dimensions, the traditional [Formula: see text]-medoids algorithm is difficult to accurately estimate the value of [Formula: see text]. Moreover, the clustering results of the average linkage (AL) algorithm cannot be divided again, and the AL algorithm cannot be directly used for text classification. In order to overcome the limitations of AL and [Formula: see text]-medoids, in this paper, we combine the two algorithms together so as to be mutually complementary to each other. In particular, in order to meet the purpose of text classification, we improve the AL algorithm and propose the [Formula: see text] testing statistics to obtain the approximate number of clusters. Finally, the central feature words are preserved, and the other feature words are deleted. The experimental results show that the new algorithm largely eliminates the redundancy of the feature. Compared with the traditional TF-IDF algorithms, the performance of the text classification of the new algorithm is improved.

A clustering of red wines based on physicochemical and optical properties

Sixty-nine samples of red wines, produced in different regions of Bulgaria, Chile and South Africa from the Merlot (42 samples) and Cabernet Sauvignon (27 samples) grape varieties are characterized by 13 physicochemical and optical properties. The physicochemical characteristics (antioxidant activity and anthocyanin content) have been obtained spectrophotometrically, while the antioxidant activity (AOA) has been measured by using both DPPH and ABTS methods. Based on a hierarchical cluster analysis with an average linkage algorithm, the studied wines from grape variety Merlot as well as the ones from grape variety Cabernet Sauvignon, according to their antioxidant activities (AOA ABTS and AOA DPPH) and anthocyanins content are grouped into three generalized groups. When integral fluorescence intensities at an excitation wavelength of 265 nm are added as a factor, for the wines from the grape variety Cabernet Sauvignon there is not any difference in their grouping, and for the ones from grape variety Merlot, the only difference is that one sample of wine passes to the first cluster. The results, obtained from the clustering procedure are presented graphically by dendrograms.

Car shape clustering using sobel edge detection with divisive average linkage and single linkage algorithm (case: bus, sedan, citycar, mpv, and truck)

This study aims to make an application that able to make clustering for car type based on shape by using edge detection and clustering algorithm. The car types being used in this study are the image of the bus, sedan, city car, mpv, and truck. In this study, the method used to classify is with the divisive average linkage and single linkage method. The goal is to prove whether the divisive average linkage and single linkage can be used to classify the shape of the car image correctly. The shape obtained by using the Sobel edge detection method because the Sobel method is an edge detector that has advantages in reducing the noise before doing calculations so that the resulting edges form an image like the original.

Assessment of ISSR Markers for Tagging Genetic Variability for Yield Components in Small Cardamom (Elettaria cardamomum Maton)

Genetic diversity among ten small cardamom accessions having variability in yield traits was investigated using inter simple sequence repeat markers (ISSR). A better understanding of the variability in yield is essential for the proper utilization of genotypes in breeding programmes. Of the 10 primers analysed, 5 reported polymorphisms and generated a total of 47 scorable loci, of which 32 were polymorphic revealing 68% polymorphism. The molecular weight of the fragments ranged from 200-1100 bp. Specific banding pattern was observed in high yielding group using ISSR7 (800bp) and with ISSR 3 (800bp) for low yielding group. Genetic diversity analysis within and among the two groups showed maximum diversity in high yielding varieties based on the values of Nei’ s genetic diversity (h) and Shannon’s informative index (I). Dendrogram based on Jaccard’s similarity coefficients were generated based on an average linkage algorithm (UPGMA) using marker data. Clear grouping of the ten accessions of small cardamom into two clusters was observed. These results suggested that ISSR markers could efficiently differentiate the small cardamom genotypes based on yield and can be useful for future cardamom improvement programmes.

Identification Bacterial Contaminant in Semiarundinaria fastuosa Tissue Culture

Bamboo is one of the plants which propagated by tissue culture technique however, the emergence of microbial contaminant caused decrease of bamboo production. A type of microbe that caused the contamination on Semiarundinaria fastuosa tissue culture is bacteria. This study aimed to isolate bacterial contaminant and understanding its biodiversity. Bacterial isolation and phenotypic characterization were done by observing morphology, physiological test, biochemical test, identification and numeric phenetic analysis. Twelve bacterial contaminants was isolated and based on profile matching with Bergey’s Manual of Determinative Bacteriology, six isolate was a member of genus Bacillus, one isolate Enterococcus, two isolate Xenorhabdus, one isolate Morococcus, one isolate Corynebacterium, and one isolate belongs to genus Sarcina. A dendrogram created using Simple Matching coefficient (SSM) parameter and average linkage algorithm shown that on the 71% similarity index, all 12 OTU was grouped in one cluster. However, a dendrogram produced by Jaccard’s coefficient (SJ) parameter and average linkage algorithm on 70% similarity index divided 12 OTU on to 12 clusters.

A simplified approach using Taqman low-density array for medulloblastoma subgrouping

Next-generation sequencing platforms are routinely used for molecular assignment due to their high impact for risk stratification and prognosis in medulloblastomas. Yet, low and middle-income countries still lack an accurate cost-effective platform to perform this allocation. TaqMan Low Density array (TLDA) assay was performed using a set of 20 genes in 92 medulloblastoma samples. The same methodology was assessed in silico using microarray data for 763 medulloblastoma samples from the GSE85217 study, which performed MB classification by a robust integrative method (Transcriptional, Methylation and cytogenetic profile). Furthermore, we validated in 11 MBs samples our proposed method by Methylation Array 450 K to assess methylation profile along with 390 MB samples (GSE109381) and copy number variations. TLDA with only 20 genes accurately assigned MB samples into WNT, SHH, Group 3 and Group 4 using Pearson distance with the average-linkage algorithm and showed concordance with molecular assignment provided by Methylation Array 450 k. Similarly, we tested this simplified set of gene signatures in 763 MB samples and we were able to recapitulate molecular assignment with an accuracy of 99.1% (SHH), 94.29% (WNT), 92.36% (Group 3) and 95.40% (Group 4), against 97.31, 97.14, 88.89 and 97.24% (respectively) with the Ward.D2 algorithm. t-SNE analysis revealed a high level of concordance (k = 4) with minor overlapping features between Group 3 and Group 4. Finally, we condensed the number of genes to 6 without significantly losing accuracy in classifying samples into SHH, WNT and non-SHH/non-WNT subgroups. Additionally, we found a relatively high frequency of WNT subgroup in our cohort, which requires further epidemiological studies. TLDA is a rapid, simple and cost-effective assay for classifying MB in low/middle income countries. A simplified method using six genes and restricting the final stratification into SHH, WNT and non-SHH/non-WNT appears to be a very interesting approach for rapid clinical decision-making.

Analysis of precipitation data in Bangladesh through hierarchical clustering and multidimensional scaling

The precipitation patterns of seventeen locations in Bangladesh from 1961 to 2014 were studied using a cluster analysis and metric multidimensional scaling. In doing so, the current research applies four major hierarchical clustering methods to precipitation in conjunction with different dissimilarity measures and metric multidimensional scaling. A variety of clustering algorithms were used to provide multiple clustering dendrograms for a mixture of distance measures. The dendrogram of pre-monsoon rainfall for the seventeen locations formed five clusters. The pre-monsoon precipitation data for the areas of Srimangal and Sylhet were located in two clusters across the combination of five dissimilarity measures and four hierarchical clustering algorithms. The single linkage algorithm with Euclidian and Manhattan distances, the average linkage algorithm with the Minkowski distance, and Ward’s linkage algorithm provided similar results with regard to monsoon precipitation. The results of the post-monsoon and winter precipitation data are shown in different types of dendrograms with disparate combinations of sub-clusters. The schematic geometrical representations of the precipitation data using metric multidimensional scaling showed that the post-monsoon rainfall of Cox’s Bazar was located far from those of the other locations. The results of a box-and-whisker plot, different clustering techniques, and metric multidimensional scaling indicated that the precipitation behaviour of Srimangal and Sylhet during the pre-monsoon season, Cox’s Bazar and Sylhet during the monsoon season, Maijdi Court and Cox’s Bazar during the post-monsoon season, and Cox’s Bazar and Khulna during the winter differed from those at other locations in Bangladesh.

The Expansion of Initial Point Algorithm for K-Modes Algorithm

The determination of the starting point in the k-modes algorithm is taken by random. Of course, such a thing can lead to an iteration of unpredictable numbers and accuracy. Therefore, it is necessary to develop different algorithms that are used to determine the starting point with hierarchical agglomerative clustering approach instead of randomly selecting the starting point in the initial iteration. At the end of this research is expected clustering process can produce more efficient iteration. The result of determining the value generated in this algorithm is the incorporation of a number of cluster central points on the variables based on the calculation of the approach which has the average linkage algorithm. Followed by calculating the difference of objective function on each iteration, of course, finished clustering process on k-modes. Iteration will stop after the difference of objective function is smaller than the specified limit.

Enhancing point symmetry-based distance for data clustering

In this paper, at first a new point symmetry-based similarity measurement is proposed which satisfies the closure and the symmetry properties of any distance function. The different desirable properties of the new distance are elaborately explained. Thereafter a new clustering algorithm based on the search capability of genetic algorithm is developed where the newly developed point symmetry-based distance is used for cluster assignment. The allocation of points to different clusters is performed in such a way that the closure property is satisfied. The proposed GA with newly developed point symmetry distance based (GAnPS) clustering algorithm is capable of determining different symmetrical shaped clusters having any sizes or convexities. The effectiveness of the proposed GAnPS clustering technique in identifying the proper partitioning is shown for twenty-one data sets having various characteristics. Performance of GAnPS is compared with existing symmetry-based genetic clustering technique, GAPS, three popular and well-known clustering techniques, K-means, expectation maximization and average linkage algorithm. In a part of the paper, the utility of the proposed clustering technique is shown for partitioning a remote sensing satellite image. The last part of the paper deals with the development of some automatic clustering techniques using the newly proposed symmetry-based distance.

Using Dendritic Heat Maps to Simultaneously Display Genotype Divergence with Phenotype Divergence.

The advancement of techniques to visualize and analyze large-scale sequencing datasets is an area of active research and is rooted in traditional techniques such as heat maps and dendrograms. We introduce dendritic heat maps that display heat map results over aligned DNA sequence clusters for a range of clustering cutoffs. Dendritic heat maps aid in visualizing the effects of group differences on clustering hierarchy and relative abundance of sampled sequences. Here, we artificially generate two separate datasets with simplified mutation and population growth procedures with GC content group separation to use as example phenotypes. In this work, we use the term phenotype to represent any feature by which groups can be separated. These sequences were clustered in a fractional identity range of 0.75 to 1.0 using agglomerative minimum-, maximum-, and average-linkage algorithms, as well as a divisive centroid-based algorithm. We demonstrate that dendritic heat maps give freedom to scrutinize specific clustering levels across a range of cutoffs, track changes in phenotype inequity across multiple levels of sequence clustering specificity, and easily visualize how deeply rooted changes in phenotype inequity are in a dataset. As genotypes diverge in sample populations, clusters are shown to break apart into smaller clusters at higher identity cutoff levels, similar to a dendrogram. Phenotype divergence, which is shown as a heat map of relative abundance bin response, may or may not follow genotype divergences. This joined view highlights the relationship between genotype and phenotype divergence for treatment groups. We discuss the minimum-, maximum-, average-, and centroid-linkage algorithm approaches to building dendritic heat maps and make a case for the divisive “top-down” centroid-based clustering methodology as being the best option visualize the effects of changing factors on clustering hierarchy and relative abundance.

A comparison of bacterial communities in feces of breastfed infants determined by 454 pyrosequencing and Illumina MiSeq (1017.5)

Because various molecular techniques are used to evaluate microbial community structure in biological samples, understanding if results are comparable among methods is critical. We hypothesized two of the most commonly‐used techniques, 454 pyrosequencing and Illumina MiSeq, similarly describe bacterial communities in breastfed infants’ feces. To test this, DNA from 25 samples was processed using both techniques; universal primers designed to amplify the V1‐V3 region of the 16S rRNA bacterial gene were used. Bacterial community profiles were clustered using Bray‐Curtis dissimilarity and the average linkage algorithm. Of the 15 most abundant genera detected in the samples, only 8 were common between the 2 methods. Diversity was similar (Shannon, P=0.08), but evenness was different between methods (Simpson, P=0.001). Our data somewhat support the hypothesis that these methods produce similar results in terms of diversity; however, the methods differ in terms of overall community composition and evenness. In conclusion, these potential differences need to be considered when comparing bacterial community data across sequencing methods ‐ at least for feces of breastfed infants.

Evaluación de la diversidad genética en genotipos de Brassica juncea (Brassicaceae) utilizando diferencias fenotípicas y marcadores SSR

Brassica mustard species represent one of the most important oilseed crops in India, nevertheless, their genetic diversity is barely known. A better understanding on this topic is essential for the proper utilization of genotypes in breeding programmes. We evaluated the genetic diversity among 44 Indian mustard (Brassica juncea) genotypes including varieties/purelines from different agro-climatic zones of India and few exotic genotypes (Australia, Poland and China). For this, we used A and B genome specific SSR markers and phenotypic data on 12 yield and yield contributing traits. Out of the 143 primers tested, 134 reported polymorphism and a total of 355 alleles were amplified. Dendrograms based on Jaccard's similarity coefficients and Manhattan dissimilarity coefficients were generated based on an average linkage algorithm (UPGMA) using marker data and phenotypic data. Genotypes were grouped into four clusters based on genetic distances. Both the clustering patterns based on Jaccard's similarity and Manhattan dissimilarity coefficients, independently, discriminated the genotypes effectively as per their pedigree and origin. PCoA revealed that, the grouping of genotypes based on SSR marker data is more convincing than phenotypic data, however, the correlation between phenotypic and genetic distance matrices was observed to be very low (r = 0.11). Hence, for diversity studies reliability on molecular markers is worth proving and SSR markers are the stronger tools than quantitative traits in discriminating B. juncea genotypes.

Principal component and clustering analysis on molecular dynamics data of the ribosomal L11·23S subdomain

With improvements in computer speed and algorithm efficiency, MD simulations are sampling larger amounts of molecular and biomolecular conformations. Being able to qualitatively and quantitatively sift these conformations into meaningful groups is a difficult and important task, especially when considering the structure-activity paradigm. Here we present a study that combines two popular techniques, principal component (PC) analysis and clustering, for revealing major conformational changes that occur in molecular dynamics (MD) simulations. Specifically, we explored how clustering different PC subspaces effects the resulting clusters versus clustering the complete trajectory data. As a case example, we used the trajectory data from an explicitly solvated simulation of a bacteria’s L11·23S ribosomal subdomain, which is a target of thiopeptide antibiotics. Clustering was performed, using K-means and average-linkage algorithms, on data involving the first two to the first five PC subspace dimensions. For the average-linkage algorithm we found that data-point membership, cluster shape, and cluster size depended on the selected PC subspace data. In contrast, K-means provided very consistent results regardless of the selected subspace. Since we present results on a single model system, generalization concerning the clustering of different PC subspaces of other molecular systems is currently premature. However, our hope is that this study illustrates a) the complexities in selecting the appropriate clustering algorithm, b) the complexities in interpreting and validating their results, and c) by combining PC analysis with subsequent clustering valuable dynamic and conformational information can be obtained.Electronic supplementary materialThe online version of this article (doi:10.1007/s00894-012-1563-4) contains supplementary material, which is available to authorized users.

Normalised monthly shortage curves: a contribution for a better understanding of monthly rain deficit in Western Europe

A new approach to the statistics of rainfall shortage at monthly scale in Western Europe is obtained from precipitation records of 115 gauges over the twentieth century. In this paper, a month is considered to have rainfall deficit when its rain amount is below the 50th percentile of the respective calendar month. The monthly shortage, MS, for every month with deficit is then computed as the absolute value of the difference between its monthly amount and the corresponding truncation level. The cumulative distributions of monthly shortage, CMS, and number of shortage months, CNM, constitute a new description of the monthly rainfall deficit. Both CMS and CNM distributions fit well to a Weibull model. Using the analogy to the normalised daily rainfall curves formulation, NRC, the relationship between CMS and CNM, named as normalised shortage curve, NSC, is modelled by the same function applied to NRCs. Similarly to NRCs, the behaviour of the NSCs strongly depends on the coefficient of variation of the monthly shortage, CVMS. Four coordinates characterising every NSC are then introduced: the CMS percentile associated with the median of CNM; the CNM percentile related to the median of CMS; and the percentiles of CMS and CNM for the average monthly shortage. In this way, the degree of asymmetric distribution of the monthly deficit is quantified. With the aim of performing a clustering process based on these four coordinates, a principal component analysis, is previously applied to remove redundancies, being obtained two uncorrelated principal components, PCs, characterising every NSC. An average linkage algorithm is then applied to these two PCs, leading to obtain spatially coherent groups of gauges with very similar NSC patterns. This clustering process permits to discard latitude and vicinity to the Atlantic Ocean or the Mediterranean Sea as main factors conditioning the monthly shortage regime.

Exploring the conformational changes of the ATP binding site of gyrase B from Escherichia coli complexed with different established inhibitors by using molecular dynamics simulation : Protein–ligand interactions in the light of the alanine scanning and free energy decomposition methods

Currently, bacterial diseases cause a death toll around 2 million people a year encouraging the search for new antimicrobial agents. DNA gyrase is a well-established antibacterial target consisting of two subunits, GyrA and GyrB, in a heterodimer A 2B 2. GyrA is involved in DNA breakage and reunion and GyrB catalyzes the hydrolysis of ATP. The GyrB subunit from Escherichia coli has been investigated, namely the ATP binding pocket both considering the protein without ligands and bound with the inhibitors clorobiocin, novobiocin and 5′-adenylyl-β-γ-imidodiphosphate. The stability of the systems was studied by molecular dynamics simulation with the further analysis of the time dependent root-mean-square coordinate deviation (RMSD) from the initial structure, and temperature factors. Moreover, exploration of the conformational space of the systems during the MD simulation was carried out by a clustering data mining technique using the average-linkage algorithm. Recognizing the key residues in the binding site of the enzyme that are involved in the binding mode with the aforementioned inhibitors was investigated by using two techniques: free energy decomposition and computational alanine scanning. The results from these simulations highlight the important residues in the ATP binding site and can be useful in the design process of potential new inhibitors.

Abscisic acid and sucrose increase the protein content in date palm somatic embryos, causing changes in 2-DE profile

Various supplements (abscisic acid (ABA) or sucrose) were added to the initial embryo culture medium (M3) with the aim of improving the vigour of vitroplants deriving from date palm somatic embryogenesis. ABA (20 and 40 μM) and sucrose (90 g/l) applied for 4 and 2 weeks respectively increased embryo thickness, with no apparent difference in length. ABA (5–40 μM) increased embryo proliferation rate. Somatic embryos maintained in modified M3 (M3 supplemented with ABA and an increased sucrose concentration) contained a higher amount of protein than those maintained in initial M3 (no ABA, 30 g/l of sucrose), with a 1.5–1.7-fold increase depending on the compound and concentration assayed. The 1-D and 2-DE protein profiles showed qualitative and quantitative differences between the somatic embryos cultured in initial M3 (control) and in modified M3. Statistical analysis of spot intensity was performed by principal component analysis, yielding two accurate groups of samples and determining the most discriminating spots. Samples were also clustered using Euclidean distance with an average linkage algorithm. Thirty-four variable spots were identified using mass spectrometry analysis. Identified proteins were classified into the following functional categories: energy metabolism (five proteins); protein translation, folding and degradation (9); redox maintenance (5); cytoskeleton (3); storage protein (2); and with no assigned function as (10). While “up-regulation” of stress-related proteins and “down-regulation” of energy metabolism proteins were observed in somatic embryos matured in M3 supplemented with ABA, storage proteins (legumin) were “up-regulated” in somatic embryos matured in M3 supplemented with increased sucrose.