Random Labeling Research Articles

Revealing rangeomorph species characters using spatial analyses

Rangeomorphs dominate the Ediacaran Avalonian macrofossil assemblages of Charnwood Forest, UK (∼562 Ma). However, their unfamiliar fractal architecture makes distinguishing phylogenetically reliable characters from intraspecific features difficult. Fortunately, spatial analysis of large in-situ populations offers an independent means of assessing their taxonomy. Populations of a single biological species are likely to exhibit similar spatial distributions due to their shared responses to the biological and ecological processes acting upon them. As such, spatial analyses can be used to interrogate which are the most taxonomically deductive characters in similar species. We used random labelling analyses to investigate the presence or absence of characters of Primocandelabrum boyntoni, P. aethelfalaedia, and P. aelfwynnia on the Bed ‘B’. The resultant spatial distributions were compared to observed characters using goodness-of-fit tests to determine which characters were associated with unique populations, and which were found across multiple populations. We found that P. boyntoni and P. aelfwynnia had statistically indistinguishable character distributions, suggesting that they represent a single biological species, and that they exhibited significantly different distributions to P. aethelfalaedia, suggesting that there are two (rather than three) species of Primocandelabrum present on the B surface. Furthermore, we found that the distribution of concealed versus unconcealed 1st order branches across all specimens exhibited significantly different density-dependant behaviour, with unconcealed branching occurring in areas of higher density populations and concealed branching occurring in the lower density areas of Primocandelabrum. We speculate that unconcealed branches may have been a response to the reduced availability of resources in higher density areas, implying rangeomorphs were capable of ecophenotypic responses.

On coincidences of tuples in a q-ary tree with random labels of vertices

Abstract Let all vertices of a complete q-ary tree of finite height be independently and equiprobably labeled by the elements of some finite alphabet. We consider the numbers of pairs of identical tuples of labels on chains of subsequent vertices in the tree. Exact formulae for the expectations of these numbers are obtained, convergence to the compound Poisson distribution is proved. For the size of cluster composed by pairs of identically labeled chains we also obtain exact formula for the expectation.

Label Stacking Scenarios in Hybrid Wavelength and Code-Switched GMPLS Networks

Multi-protocol label switching (MPLS) is a promising solution to implement high-speed internet protocol (IP) networks by reducing the layer number. To meet the increasing demand for data traffic, optical packet switching (OPS) is integrated under IP to provide high bandwidth to end users. Generalized MPLS (GMPLS) is perfectly compatible with the routing algorithm in IP/MPLS as it supports packet-switching functions. In this paper, we investigate the label stacking scenarios in GMPLS networks. In GMPLS, label stacking is done to reduce the node complexity by appending multiple labels to a single packet. Wavelength-division multiplexing (WDM) and optical code-division multiplexing (OCDM) signals have been widely used as identifying labels. As the labels can be permutated among the wavelengths or code dimensions, the structure of a label stack can be varied. However, studies on the relationship between label stacking scenarios and network performance are limited. To investigate this issue, we propose three label stacking models: sequential code distribution; sequential wavelength distribution, and random label distribution. The simulation results show that the sequential wavelength assignment, wherein the labels are uniformly distributed among the wavelengths, exhibits the best system performance in terms of the label-error rate (LER).

Conformational Transitions of Polymer Chains in Solutions Characterized by Fluorescence Resonance Energy Transfer.

The critical overlap concentration C* is an important concept in polymer solutions and is defined as the boundary between dilute and semidilute regimes. In this study, the chain conformational changes of polystyrene (PS) with both high (Mn = 200,000 Da) and low (Mn = 13,000 Da) molecular weights in cis-decalin were compared by intrachain fluorescence resonance energy transfer (FRET). The random labeling of donor and acceptor chromophores strategy was employed for long PS chains, whereas chain-end labeling was used for short PS chains. By monitoring the spectroscopic intensity ratio between acceptor and donor, the concentration dependence on chain conformation from dilute to semidilute solutions was determined. Both long and short chains exhibit a conformational transition concentration, above which the polymer chains begin to collapse with concentration significantly. Interestingly, for randomly labeled polymer long chains, such concentration is consistent with C* determined from the viscosity result, below which only slight conformational change of polymer chain takes place. However, for the chain-end labeled short chain, the conformational transition concentration takes place earlier than C*, below which no significant polymer conformation change is observed.

Characterization of Fluorescently Labeled Protein with Electrospray Ionization-MS and Fluorescence Spectroscopy: How Random is Random Labeling?

Solvent exposed lysine residues are abundantly present in many proteins. Their highly reactive ε-amino groups serve as universal targets for coupling with active esters of various extrinsic labels including a vast arsenal of fluorescent probes. Here, we describe fluorescent properties and preferential localization of two frequently used fluorescent labels, AlexaFluor488 (AF488) and Cy3, on the surface of a small highly soluble serum protein neutrophil gelatinase-associated lipocalin (NGAL), which serves as a diagnostic marker for acute kidney failure. Using a standard protocol for labeling with either AF488-SDP or AF488-NHS, we achieved >95% labeling efficiency of the protein as determined by UV-vis absorption and electrospray ionization (ESI)-MS. However, fluorescence intensity of the labeled protein was less than 10% of the expected value. To understand the unusually high quenching of the probe, we identified the sites of AF488 attachments by means of LC-MS/MS combined with trypsin digestion. Surprisingly, we found that the AF488 label is not randomly distributed among accessible lysines but predominantly bound to the residues K125, K126, or K135, which are located in the NGAL calyx and are likely quenched by neighboring tryptophans and tyrosines. In contrast, when NGAL was labeled with Cy3, the probe's fluorescence was almost fully retained. The LC-MS/MS data indicated that Cy3 was predominately bound to another lysine, K31, on the protein surface on the opposite side of the calyx. Our findings suggest that a combination of the inherent properties of the label and the specifics of the protein microenvironment may selectively lead probes to specific lysine residues and thus challenge the common view that protein labeling is a random process.

Active learning based semi-supervised exponential discriminant analysis and its application for fault classification in industrial processes

For the industrial fault classification, exponential discriminant analysis (EDA) requires that all the training samples should be labeled; however, only a minority of the training samples are randomly labeled in real industrial processes. This motivates the formulation of the active learning based semi-supervised exponential discriminant analysis in this paper. Firstly, to make EDA applicable to the semi-supervised industrial scenario, scatter matrices are transformed into their regularization variants through combining unlabeled training samples. Moreover, to reduce the adverse effect of random labeling of training samples, the best versus second-best rule is employed to select more informative training samples in an active way to upgrade the model classification performance. And the obvious performance improvement of the proposed method is demonstrated with extensive experiments on synthesized data, the TE process and a real industrial process.

Overoptimism in cross-validation when using partial least squares-discriminant analysis for omics data: a systematic study

Advances in analytical instrumentation have provided the possibility of examining thousands of genes, peptides, or metabolites in parallel. However, the cost and time-consuming data acquisition process causes a generalized lack of samples. From a data analysis perspective, omics data are characterized by high dimensionality and small sample counts. In many scenarios, the analytical aim is to differentiate between two different conditions or classes combining an analytical method plus a tailored qualitative predictive model using available examples collected in a dataset. For this purpose, partial least squares-discriminant analysis (PLS-DA) is frequently employed in omics research. Recently, there has been growing concern about the uncritical use of this method, since it is prone to overfitting and may aggravate problems of false discoveries. In many applications involving a small number of subjects or samples, predictive model performance estimation is only based on cross-validation (CV) results with a strong preference for reporting results using leave one out (LOO). The combination of PLS-DA for high dimensionality data and small sample conditions, together with a weak validation methodology is a recipe for unreliable estimations of model performance. In this work, we present a systematic study about the impact of the dataset size, the dimensionality, and the CV technique used on PLS-DA overoptimism when performance estimation is done in cross-validation. Firstly, by using synthetic data generated from a same probability distribution and with assigned random binary labels, we have obtained a dataset where the true classification rate (CR) is 50%. As expected, our results confirm that internal validation provides overoptimistic estimations of the classification accuracy (i.e., overfitting). We have characterized the CR estimator in terms of bias and variance depending on the internal CV technique used and sample to dimensionality ratio. In small sample conditions, due to the large bias and variance of the estimator, the occurrence of extremely good CRs is common. We have found that overfitting peaks when the sample size in the training subset approaches the feature vector dimensionality minus one. In these conditions, the models are neither under- or overdetermined with a unique solution. This effect is particularly intense for LOO and peaks higher in small sample conditions. Overoptimism is decreased beyond this point where the abundance of noisy produces a regularization effect leading to less complex models. In terms of overfitting, our study ranks CV methods as follows: Bootstrap produces the most accurate estimator of the CR, followed by bootstrapped Latin partitions, random subsampling, K-Fold, and finally, the very popular LOO provides the worst results. Simulation results are further confirmed in real datasets from mass spectrometry and microarrays.

Random forest for label ranking

Label ranking aims to learn a mapping from instances to rankings over a finite number of predefined labels. Random forest is a powerful and one of the most successful general-purpose machine learning algorithms of modern times. In this paper, we present a powerful random forest label ranking method which uses random decision trees to retrieve nearest neighbors. We have developed a novel two-step rank aggregation strategy to effectively aggregate neighboring rankings discovered by the random forest into a final predicted ranking. Compared with existing methods, the new random forest method has many advantages including its intrinsically scalable tree data structure, highly parallel-able computational architecture and much superior performance. We present extensive experimental results to demonstrate that our new method achieves the highly competitive performance compared with state-of-the-art methods for datasets with complete ranking and datasets with only partial ranking information.

Linear Chains of HER2 Receptors Found in the Plasma Membrane Using Liquid-Phase Electron Microscopy

The spatial distribution of the human epidermal growth factor 2 (HER2) receptor in the plasma membrane of SKBR3 and HCC1954 breast cancer cells was studied. The receptor was labeled with quantum dot nanoparticles, and fixed whole cells were imaged in their native liquid state with environmental scanning electron microscopy using scanning transmission electron microscopy detection. The locations of individual HER2 positions were determined in a total plasma membrane area of 991 μm2 for several SKBR3 cells and 1062 μm2 for HCC1954 cells. Some of the HER2 receptors were arranged in a linear chain with interlabel distances of 40 ± 7 and 32 ± 10 nm in SKBR3 and HCC1954 cells, respectively. The finding was tested against randomly occurring linear chains of six or more positions, from which it was concluded that the experimental finding is significant and did not arise from random label distributions. Because the measured interlabel distance in the HER2 chains is similar to the 36-nm helix-repetition distance of actin filaments, it is proposed that a linking mechanism between HER2 and actin filaments leads to linearly aligned oligomers.

Grouping Objects to Homogeneous Classes Satisfying Requisite Mass

Grouping datasets plays an important role in many scientific researches. Depending on data features and applications, different constrains are imposed on groups, while having groups with similar members is always a main criterion. In this paper, we propose an algorithm for grouping the objects with random labels, nominal features having too many nominal attributes. In addition, the size constraint on groups is necessary. These conditions lead to a mixed integer optimization problem which is not convex nor linear. It is an NP-hard problem and exact solution methods are computationally costly. Our motivation to solve such a problem comes along with grouping insurance data which is essential for fair pricing. The proposed algorithm includes two phases. First, we rank random labels using fuzzy numbers. Afterwards, an adjusted K-means algorithm is used to produce homogenous groups satisfying a cluster size constraint. Fuzzy numbers are used to compare random labels, in both observed values and their chance of occurrence. Moreover, an index is defined to find the similarity of multi-valued attributes without perfect information with those accompanied with perfect information. Since all ranks are scaled into the interval [0,1], the result of ranking random labels does not need rescaling techniques. In the adjusted K-means algorithm, the optimum number of clusters is found using coefficient of variation instead of Euclidean distance. Experiments demonstrate that our proposed algorithm produces fairly homogenous and significantly different groups having requisite mass.

A multidimensional spatial scan statistics approach to movement pattern comparison

ABSTRACTThis paper describes a multidimensional spatial scan statistics approach to comparing spatial movement patterns based on origin–destination (OD) representation. This approach aims to evaluate differences and similarities between the spatial distributions of a pair of OD movement datasets, and detect areas where the two spatial distributions differ the most. Specifically, two OD datasets being compared are modeled as a bivariate marked spatial point process in a multidimensional space, consisting of points representing individual OD movement records. Such multidimensional space is formed by the Cartesian product of the origins’ and the destinations’ geographic spaces. With this spatial data model, one can evaluate how two movement distributions differ from each other by testing against a random labeling null hypothesis. A multidimensional Bernoulli spatial scan statistics method is developed to detect OD region pairs with abnormally high concentrations of one movement dataset over the other. The existence and the spatial extents of these OD region pairs indicate whether and where the two movement distributions differ. Two case studies were conducted to evaluate the approach by comparing morning and afternoon taxi trips (individual movements), and county-to-county migration flows between age groups (aggregated movement flows), and demonstrated that areas with the most significant spatial distribution differences could be detected from large movement datasets.

Bearing Fault Automatic Classification Based on Deep Learning

An automatic classification method based on deep learning for bearing fault diagnosis is proposed. The method is designed with the ability of faulty signal automatic clustering without human knowledge. A dataset in which each sample is given a random label is configured after extracting the features of vibration signals from the frequency domain. The dataset is used to train a deep neural network (DNN) to obtain the initial classification. The classification results are assessed by testing the subsignals extracted from the raw data, and the sample labels are modified according to the evaluation result. The modified dataset is used to train the DNN a second time. Samples with characteristic faults are clustered in various classes after iterating the DNN training and testing. The proposed method is tested with the bearing data provided by the Case Western Reserve University (CWRU) Bearing Data Center, which is a standard reference to test fault detection algorithms. The 12k drive end, 48k drive end, and 12k fan end CWRU bearing data are classified into 7, 6, and 4 groups, respectively. The testing results show that the proposed method can achieve automatic clustering for vibration signals with a variety of faults.

Iterative image segmentation with feature driven heuristic four-color labeling

Multilabel segmentation is an important research branch in image segmentation field. In our previous work, Multiphase Multiple Piecewise Constant and Geodesic Active Contour (MMPC-GAC) model was proposed, which can effectively describe multiple objects and background with intensity inhomogeneity. It can be approximately solved with Multiple Layer Graph (MLG) methods. To make the optimization more efficient and limit the approximate error, four-color labeling theorem was further introduced which can limit the MLG within three layers (representing four phases). However, the adopted random four-color labeling method usually provides chaotic color maps with obvious inhomogeneity for those semantic consistent regions. For this case, a new and alternative method named heuristic four-color labeling is proposed in this paper, which aims to generate more reasonable color maps with a global view of the whole image. And compared with the random four-color labeling strategy, the whole iterative algorithm based on our method usually produces better segmentations with faster convergence, particularly for images with clutters and complicated structures. This strategy is a good substitute for random coloring method when the latter produces unsatisfactory messy segmentation. Experiments conducted on public dataset demonstrate the effectiveness of the proposed method.

CRISPR/Cas9 screening using unique molecular identifiers

Loss‐of‐function screening by CRISPR/Cas9 gene knockout with pooled, lentiviral guide libraries is a widely applicable method for systematic identification of genes contributing to diverse cellular phenotypes. Here, Random Sequence Labels (RSLs) are incorporated into the guide library, which act as unique molecular identifiers (UMIs) to allow massively parallel lineage tracing and lineage dropout screening. RSLs greatly improve the reproducibility of results by increasing both the precision and the accuracy of screens. They reduce the number of cells needed to reach a set statistical power, or allow a more robust screen using the same number of cells.

Spatial patterns of lost and remaining trees in the Iberian wooded rangelands

Spanish dehesas and Portuguese montados are similar Iberian silvopastoral systems that provide important environmental services and hold a high biodiversity. However, there is a raising concern for the sustainability of these cultural landscapes, as they suffer from tree recruitment failure. Consequently, trees in those systems are lessening due to tree ageing, diseases and other causes. A key to developing successful management strategies in dehesas is to understand the spatiotemporal behaviour of this process. We analysed the spatial pattern of lost and remaining trees in five dehesas in Extremadura (SW Spain) between 1956 and 2009 based on aerial images. We studied the possible spatial dependency (clustering or segregation) of the lost and remaining trees in areas with low or high land use intensity at different spatial distances. Analyses were performed under the random labelling hypothesis and the critical values were obtained by using 199 permutations of a Monte Carlo test. A linear mixed model was used to investigate the possible factors affecting tree loss. The results indicate a clear segregation between remaining and lost trees in areas of high intensity land use, while in areas with low intensity this pattern was less evident. Also, lost trees were more clustered than the remaining ones. Higher tree loss density occurred at lower slopes, the areas initially occupied by high or medium-density shrubs being less prone to undergo tree loss than those that appeared in 1956 like ploughed lands, grasslands or low-density shrubs. These results show a concentration of the risk of tree loss in certain areas where conservation efforts should be focused.

Abstract 2858: Immuno-PET with site-specific labeled 89Zr-DFO-trastuzumab improves immuno-reactivity and tumor uptake in a subcutaneous HER2 positive xenograft mouse model of ovarian adenocarcinoma

Abstract Introduction: The combination of PET with the specificity of monoclonal antibodies (mAbs), immuno-PET, is an attractive approach to improve tumor detection and mAb quantification. 89Zirconium (89Zr) is particularly well-matched for the long circulatory half-life of intact mAbs. Previously, the use of 89Zr has been limited by the lack of suitable methods for stable coupling to mAbs. In addition, the conventional labeling strategies with random introduction of a chelator potentially reduce antibody binding and affinity towards target. Here we demonstrate the application of an enzyme- and click chemistry-mediated methodology for site-specifically labeled 89Zr immuno-PET imaging probes and compare them to a random labeled probe. Experimental procedures: Trastuzumab was conjugated to the p-SCN-Bn-Deferoxamine (p-SCN-Bn-DFO) chelator in three alternate ways: (1) randomly on lysine residues or site-specifically on enzymatically treated glycans using either (2) ß-galactosidase or (3) endoglycosidase S2. DFO-trastuzumab was radiolabeled with 89Zr (89Zr-DFO-trastuzumab) and injected into SK-OV-3 tumor-bearing NMRI nude mice. In addition, a dose-escalation study was performed with co-injection of either 100 µg or 500 µg of unlabeled trastuzumab. At 24, 70 and 110 hours post-injection, mice underwent small animal PET/CT imaging for longitudinal assessment. Mice were euthanized and organs resected for ex vivo biodistribution after the last imaging time-point. Additionally, the immuno-reactivity and tracer stability in buffer and plasma were assessed for all three tracers. Results: All tracers were found to be stable for up to 5 days post-labeling. PET imaging at 24, 70 and 110 hours after injection of 89Zr-DFO-trastuzumab revealed a gradual increase in tumor uptake and image contrast over the time-course. At 70 hours post-injection mean tumor uptakes were 6.7 ± 1.7, 13.9 ± 3.3 and 15.3 ± 3.8 % injected dose per gram tissue (%ID/g), for random, ß-galactosidase or endoglycosidase S2 labeled probes, respectively. Site-specific labeling significantly increased tumor PET uptake (One-way ANOVA, p < 0.0001) at all time-points compared to random labeling. This was further supported by an immuno-reactivity of 93% for site-specific labeled probes compared to 80% for the random labeled trastuzumab. Titration with unlabeled trastuzumab did not affect tumor uptake considerably and the ex vivo biodistribution confirmed the data obtained by in vivo PET imaging. Conclusions: 89Zr-DFO-trastuzumab is well suited for specific immuno-PET imaging of HER2 positive ovarian cancer and site-specific labeling of trastuzumab presents with markedly higher immuno-reactivity and tumor specificity. These findings support the further development of site-specific radiolabeled monoclonal antibodies for immuno-PET. Citation Format: Lotte K. Kristensen, Camilla Christensen, Camilla S. Knudsen, Mette M. Jensen, Brian J. Agnew, Andreas Kjaer, Carsten H. Nielsen. Immuno-PET with site-specific labeled 89Zr-DFO-trastuzumab improves immuno-reactivity and tumor uptake in a subcutaneous HER2 positive xenograft mouse model of ovarian adenocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 2858. doi:10.1158/1538-7445.AM2017-2858

Age-structured population model of cell survival.

Age-structured cell population model was introduced to describe cell survival. The impact of the environment on the cell population is represented by drug plasma concentration. A key model variable is the hazard of cell removal that is a subject to the environment effect. The model is capable of describing cohort and random labeling cell survival data. In addition, it accounts for cell loss due to labeling of cell sample, but it lacks ability to describe the effect of label elution on the survival data. The model was applied to red blood cell (RBC) survival data in two groups of Wistar rats obtained by two techniques: cohort labeling using 14C-glycine (N=4) and random labeling using biotin (N=8). The Weibull probability density function was selected for the RBC lifespan distribution. The data were simultaneously fitted by the mixed effects model implemented in Monolix 4.3.3. The estimated typical values of RBC lifespan and age were 53.7 and 27.8days, respectively. A noticeable effect of biotinylation on RBC survival was observed that resulted in a significant difference between the means of individual RBC lifespan for two groups. The model provides a mechanistic framework flexible enough to account for various experimental designs to generate the cell survival data. Despite model qualification using animal data, the model has the same potential to be applied to cell survival data analysis in humans.

Effect of Training Class Label Noise on Classification Performances for Land Cover Mapping with Satellite Image Time Series

Supervised classification systems used for land cover mapping require accurate reference databases. These reference data come generally from different sources such as field measurements, thematic maps, or aerial photographs. Due to misregistration, update delay, or land cover complexity, they may contain class label noise, i.e., a wrong label assignment. This study aims at evaluating the impact of mislabeled training data on classification performances for land cover mapping. Particularly, it addresses the random and systematic label noise problem for the classification of high resolution satellite image time series. Experiments are carried out on synthetic and real datasets with two traditional classifiers: Support Vector Machines (SVM) and Random Forests (RF). A synthetic dataset has been designed for this study, simulating vegetation profiles over one year. The real dataset is composed of Landsat-8 and SPOT-4 images acquired during one year in the south of France. The results show that both classifiers are little influenced for low random noise levels up to 25%–30%, but their performances drop down for higher noise levels. Different classification configurations are tested by increasing the number of classes, using different input feature vectors, and changing the number of training instances. Algorithm complexities are also analyzed. The RF classifier achieves high robustness to random and systematic label noise for all the tested configurations; whereas the SVM classifier is more sensitive to the kernel choice and to the input feature vectors. Finally, this work reveals that the cross-validation procedure is impacted by the presence of class label noise.

A UV cross-linking method combined with infrared imaging to analyse RNA-protein interactions.

Photo cross-linking of proteins with short RNA oligomers is a classical method to study RNA–protein interactions that are implicated in many aspects of RNA metabolism and function. Most commonly, this involves the use of [γ-32P]-labeled RNA probes. Although very sensitive, these procedures are complicated by the safety issues associated with the use of radioisotopes. Here, we describe a modified UV cross-linking method using oligonucleotide probes end labelled with the infrared dye IRDye®800. After UV cross-linking, proteins are separated by SDS-PAGE and cross-linked products are visualized with the Odyssey® Infrared Imaging system. This end labelling approach provides a streamlined alternative to random labelling which reduces the efficiency of in-vitro transcription. End labelling is also independent of the length of the probe, thus facilitating quantitative comparisons. To validate the method, we have confirmed the binding of HuD to the 3′-UTR of the mRNA for the microtubule-associated protein tau, implicated in the pathogenesis of Alzheimer’s disease. UV cross-linking of HuD with a labeled 21-mer probe was successfully performed using a recombinant purified glutathione-S-transferase–HuD fusion protein as well as with lysates from CHO cells transfected with HuD cDNA. UV cross-linking combined with infrared imaging offers a convenient and robust strategy to analyse RNA–protein interactions and their emerging importance in disease.

Spatial patterns of an endemic Mediterranean palm recolonizing old fields.

Throughout Europe, increased levels of land abandonment lead to (re)colonization of old lands by forests and shrublands. Very little is known about the spatial pattern of plants recolonizing such old fields. We mapped in two 21–22‐ha plots, located in the Doñana National Park (Spain), all adult individuals of the endozoochorous dwarf palm Chamaerops humilis L. and determined their sex and sizes. We used techniques of spatial point pattern analysis (SPPA) to precisely quantify the spatial structure of these C. humilis populations. The objective was to identify potential processes generating the patterns and their likely consequences on palm reproductive success. We used (1) Thomas point process models to describe the clustering of the populations, (2) random labeling to test the sexual spatial segregation, and (3) mark correlation functions to assess spatial structure in plant sizes. Plants in both plots showed two critical scales of clustering, with small clusters of a radius of 2.8–4 m nested within large clusters with 38–44 m radius. Additional to the clustered individuals, 11% and 27% of all C. humilis individuals belonged to a random pattern that was independently superimposed to the clustered pattern. The complex spatial pattern of C. humilis could be explained by the effect of different seed‐dispersers and predators' behavior and their relative abundances. Plant sexes had no spatial segregation. Plant sizes showed a spatial aggregation inside the clusters, with a decreasing correlation with distance. Clustering of C. humilis is strongly reliant on its seed dispersers and stressful environmental conditions. However, it seems that the spatial patterns and dispersal strategies of the dwarf palm make it a successful plant for new habitat colonization. Our results provide new information on the colonization ability of C. humilis and can help to develop management strategies to recover plant populations.