DNA methylation is an epigenetic modification involved in genomic stability, gene regulation, development and disease. DNA methylation occurs mainly through the addition of a methyl group to cytosines, for example to cytosines in a CpG dinucleotide context (CpG stands for a cytosine followed by a guanine). Tissue-specific methylation patterns lead to genomic regions with different characteristic methylation levels. E.g. in vertebrates CpG islands (regions with high CpG content) that are associated to promoter regions of expressed genes tend to be unmethylated. MethEvolSIM is a model-based simulation software for the generation and modification of cytosine methylation patterns along a given tree, which can be a genealogy of cells within an organism, a coalescent tree of DNA sequences sampled from a population, or a species tree. The simulations are based on an extension of the model of Grosser & Metzler (2020) <doi:10.1186/s12859-020-3438-5> and allows for changes of the methylation states at single cytosine positions as well as simultaneous changes of methylation frequencies in genomic structures like CpG islands.

This package provides a new S3 class object and relevant methods to analyze biogas fermentation data. It includes three workflows. One is specialized to a commercially available lab-scale fermentation system (see e.g. Nwaigwe (2018) <doi:10.1115/ES2018-7553>). The second provides more flexibility and allows to import data from plain text files. The last workflow offers the most flexibility as it doesn't expect external input files but relays on interactive user input. Although the focus is set on biogas fermentations, concepts and workflows may be also applicable to other fermentations even if not a gaseous product is measured. Furthermore, it provides functions that bridge to established plot engines (e.g. ggplot2 or plotly') for data visualisation. bgfanalyzer catches up an idea of Hafner et al. (2018) <doi:10.1016/j.softx.2018.06.005> of using R to standardize research within the biogas field. For more details on standardization efforts within the biogas research field see Hollinger et al. (2016) <doi:10.2166/wst.2016.336> and Hollinger et al. (2021) <doi:10.2166/wst.2020.569>.

The differences in the RNA types being sequenced have an impact on the resulting sequencing profiles. mRNA-seq data is enriched with reads derived from exons, while GRO-, nucRNA- and chrRNA-seq demonstrate a substantial broader coverage of both exonic and intronic regions. The presence of intronic reads in GRO-seq type of data makes it possible to use it to computationally identify and quantify all de novo continuous regions of transcription distributed across the genome. This type of data, however, is more challenging to interpret and less common practice compared to mRNA-seq. One of the challenges for primary transcript detection concerns the simultaneous transcription of closely spaced genes, which needs to be properly divided into individually transcribed units. The R package transcriptR combines RNA-seq data with ChIP-seq data of histone modifications that mark active Transcription Start Sites (TSSs), such as, H3K4me3 or H3K9/14Ac to overcome this challenge. The advantage of this approach over the use of, for example, gene annotations is that this approach is data driven and therefore able to deal also with novel and case specific events.

Assists in the plotting and functional smoothing of traits measured over time and the extraction of features from these traits, implementing the SET (Smoothing and Extraction of Traits) method described in Brien et al. (2020) Plant Methods, 16. Smoothing of growth trends for individual plants using natural cubic smoothing splines or P-splines is available for removing transient effects and segmented smoothing is available to deal with discontinuities in growth trends. There are graphical tools for assessing the adequacy of trait smoothing, both when using this and other packages, such as those that fit nonlinear growth models. A range of per-unit (plant, pot, plot) growth traits or features can be extracted from the data, including single time points, interval growth rates and other growth statistics, such as maximum growth or days to maximum growth. The package also has tools adapted to inputting data from high-throughput phenotyping facilities, such from a Lemna-Tec Scananalyzer 3D (see <https://www.youtube.com/watch?v=MRAF_mAEa7E/> for more information). The package growthPheno can also be installed from <http://chris.brien.name/rpackages/>.

Enhancing cross-language compatibility within the RStudio environment and supporting seamless language understanding, the deepRstudio package leverages the power of the DeepL API (see <https://www.deepl.com/docs-api>) to enable seamless, fast, accurate, and affordable translation of code comments, documents, and text. This package offers the ability to translate selected text into English (EN), as well as from English into various languages, namely Japanese (JA), Chinese (ZH), Spanish (ES), French (FR), Russian (RU), Portuguese (PT), and Indonesian (ID). With much of the text being written in English, the emphasis is on compatibility from English. It is also designed for developers working on multilingual projects and data analysts collaborating with international teams, simplifying the translation process and making code more accessible and comprehensible to people with diverse language backgrounds. This package uses the rstudioapi package and DeepL API, and is simply implemented, executed from addins or via shortcuts on RStudio'. With just a few steps, content can be translated between supported languages, promoting better collaboration and expanding the global reach of work. The functionality of this package works only on RStudio using rstudioapi'.

Data quality assessment is an integral part of preparatory data analysis to ensure sound biological information retrieval. We present here the MatrixQCvis package, which provides shiny-based interactive visualization of data quality metrics at the per-sample and per-feature level. It is broadly applicable to quantitative omics data types that come in matrix-like format (features x samples). It enables the detection of low-quality samples, drifts, outliers and batch effects in data sets. Visualizations include amongst others bar- and violin plots of the (count/intensity) values, mean vs standard deviation plots, MA plots, empirical cumulative distribution function (ECDF) plots, visualizations of the distances between samples, and multiple types of dimension reduction plots. Furthermore, MatrixQCvis allows for differential expression analysis based on the limma (moderated t-tests) and proDA (Wald tests) packages. MatrixQCvis builds upon the popular Bioconductor SummarizedExperiment S4 class and enables thus the facile integration into existing workflows. The package is especially tailored towards metabolomics and proteomics mass spectrometry data, but also allows to assess the data quality of other data types that can be represented in a SummarizedExperiment object.

Conduct numerous exploratory analyses in an instant with a point-and-click interface. With one simple command, this tool launches a Shiny App on the local machine. Drag and drop variables in a data set to categorize them as possible independent, dependent, moderating, or mediating variables. Then run dozens (or hundreds) of analyses instantly to uncover any statistically significant relationships among variables. Any relationship thus uncovered should be tested in follow-up studies. This tool is designed only to facilitate exploratory analyses and should NEVER be used for p-hacking. Many of the functions used in this package are previous versions of functions in the R Packages kim and ezr'. Selected References: Chang et al. (2021) <https://CRAN.R-project.org/package=shiny>. Dowle et al. (2021) <https://CRAN.R-project.org/package=data.table>. Kim (2023) <https://jinkim.science/docs/kim.pdf>. Kim (2021) <doi:10.5281/zenodo.4619237>. Kim (2020) <https://CRAN.R-project.org/package=ezr>. Simmons et al. (2011) <doi:10.1177/0956797611417632> Tingley et al. (2019) <https://CRAN.R-project.org/package=mediation>. Wickham et al. (2020) <https://CRAN.R-project.org/package=ggplot2>.

This package provides clean, tidy access to the Anthropic Economic Index (AEI) dataset hosted on Hugging Face <https://huggingface.co/datasets/Anthropic/EconomicIndex>. The AEI is a recurring release from Anthropic that maps usage of the Claude family of large language models to occupations and tasks using the O*NET taxonomy and the Standard Occupational Classification system, following the methodology of Handa et al. (2025) <doi:10.48550/arXiv.2503.04761> and the privacy-preserving system Clio of Tamkin et al. (2024) <doi:10.48550/arXiv.2412.13678>. Functions list available releases, fetch raw and enriched usage tables, retrieve task statements, request hierarchies, country-level breakdowns, and the standalone labor market impacts tables (job exposure and task penetration), compare two releases, join the index to user-supplied data on a shared key, and compute usage-concentration metrics (Herfindahl-Hirschman Index, top-N concentration ratios, Shannon entropy). Data is cached locally for subsequent calls. Reproducibility helpers produce BibTeX or plain-text citations that include the methodological source paper. This product uses the Anthropic Economic Index data (released under CC-BY by Anthropic') but is not endorsed or certified by Anthropic'.

Calculates the chilling and heat accumulation for studies of the temperate fruit trees. The models in this package are: Utah (Richardson et al., 1974, ISSN:0018-5345), Positive Chill Units - PCU (Linsley-Noaks et al., 1995, ISSN:1017-0316), GDH-A - Growing Degree Hours by Anderson et al.(1986, ISSN:0567-7572), GDH-R - Growing Degree Hours by Richardson et al.(1975, ISSN:0018-5345), North Carolina (Shaltout e Unrath, 1983, ISSN:0003-1062), Landsberg Model (Landsberg, 1974, ISSN:0305-7364), Q10 Model (Bidabe, 1967, ISSN:0031-9368), Jones Model (Jones et al., 2013 <DOI:10.1111/j.1438-8677.2012.00590.x>), Low-Chill Model (Gilreath and Buchanan, 1981, ISSN:0003-1062), Model for Cherry "Sweetheart" (Guak and Nielsen, 2013 <DOI:10.1007/s13580-013-0140-9>), Model for apple "Gala" (Guak and Nielsen, 2013 <DOI:10.1007/s13580-013-0140-9>), Taiwan Model (Lu et al., 2012 <DOI:10.17660/ActaHortic.2012.962.35>), Dynamic Model (Fishman et al., 1987, ISSN:0022-5193) adapted from the function Dynamic_Model() of the chillR package (Luedeling, 2018), Unified Model (Chuine et al., 2016 <DOI:10.1111/gcb.13383>) and Heat Restriction model.

The heterogeneity of spatial data presenting a finite number of categories can be measured via computation of spatial entropy. Functions are available for the computation of the main entropy and spatial entropy measures in the literature. They include the traditional version of Shannon's entropy (Shannon, 1948 <doi:10.1002/j.1538-7305.1948.tb01338.x>), Batty's spatial entropy (Batty, 1974 <doi:10.1111/j.1538-4632.1974.tb01014.x>), O'Neill's entropy (O'Neill et al., 1998 <doi:10.1007/BF00162741>), Li and Reynolds contagion index (Li and Reynolds, 1993 <doi:10.1007/BF00125347>), Karlstrom and Ceccato's entropy (Karlstrom and Ceccato, 2002 <https://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-61351>), Leibovici's entropy (Leibovici, 2009 <doi:10.1007/978-3-642-03832-7_24>), Parresol and Edwards entropy (Parresol and Edwards, 2014 <doi:10.3390/e16041842>) and Altieri's entropy (Altieri et al., 2018, <doi:10.1007/s10651-017-0383-1>). Full references for all measures can be found under the topic SpatEntropy'. The package is able to work with lattice and point data. The updated version works with the updated spatstat package (>= 3.0-2).

This package provides a random forest variant block forest ('BlockForest') tailored to the prediction of binary, survival and continuous outcomes using block-structured covariate data, for example, clinical covariates plus measurements of a certain omics data type or multi-omics data, that is, data for which measurements of different types of omics data and/or clinical data for each patient exist. Examples of different omics data types include gene expression measurements, mutation data and copy number variation measurements. Block forest are presented in Hornung & Wright (2019). The package includes four other random forest variants for multi-omics data: RandomBlock', BlockVarSel', VarProb', and SplitWeights'. These were also considered in Hornung & Wright (2019), but performed worse than block forest in their comparison study based on 20 real multi-omics data sets. Therefore, we recommend to use block forest ('BlockForest') in applications. The other random forest variants can, however, be consulted for academic purposes, for example, in the context of further methodological developments. Reference: Hornung, R. & Wright, M. N. (2019) Block Forests: random forests for blocks of clinical and omics covariate data. BMC Bioinformatics 20:358. <doi:10.1186/s12859-019-2942-y>.

When causal quantities are not identifiable from the observed data, it still may be possible to bound these quantities using the observed data. We outline a class of problems for which the derivation of tight bounds is always a linear programming problem and can therefore, at least theoretically, be solved using a symbolic linear optimizer. We extend and generalize the approach of Balke and Pearl (1994) <doi:10.1016/B978-1-55860-332-5.50011-0> and we provide a user friendly graphical interface for setting up such problems via directed acyclic graphs (DAG), which only allow for problems within this class to be depicted. The user can then define linear constraints to further refine their assumptions to meet their specific problem, and then specify a causal query using a text interface. The program converts this user defined DAG, query, and constraints, and returns tight bounds. The bounds can be converted to R functions to evaluate them for specific datasets, and to latex code for publication. The methods and proofs of tightness and validity of the bounds are described in a paper by Sachs, Jonzon, Gabriel, and Sjölander (2022) <doi:10.1080/10618600.2022.2071905>.

Features include the ability to extract tabled content from NISO-JATS-coded XML, any native HTML or HML file, DOCX, and PDF documents, and then collapse it into a text format that is readable by humans by mimicking the actions of a screen reader. As tables within PDF documents are extracted with the tabulapdf package, and the table captions and footnotes cannot be extracted, the results on tables within PDF documents have to be considered less precise. The function table2matrix() returns a list of the tables within a document as character matrices. table2text() collapses the matrix content into a list of character strings by imitating the behavior of a screen reader. The textual representation of characters and numbers can be unified with unifyMatrix() before parsing. The function table2stats() extracts the tabled statistical test results from the collapsed text with the function standardStats() from the JATSdecoder package and, if activated, checks the reported and coded p-values for consistency. Due to the great variability and potential complexity of table structures, parsing accuracy may vary. A detailed description of how tableParser works is provided here: <doi:10.48550/arXiv.2603.19756>.

Computes various entropies of given time series. This is the initial version that includes ApEn() and SampEn() functions for calculating approximate entropy and sample entropy. Approximate entropy was proposed by S.M. Pincus in "Approximate entropy as a measure of system complexity", Proceedings of the National Academy of Sciences of the United States of America, 88, 2297-2301 (March 1991). Sample entropy was proposed by J. S. Richman and J. R. Moorman in "Physiological time-series analysis using approximate entropy and sample entropy", American Journal of Physiology, Heart and Circulatory Physiology, 278, 2039-2049 (June 2000). This package also contains FastApEn() and FastSampEn() functions for calculating fast approximate entropy and fast sample entropy. These are newly designed very fast algorithms, resulting from the modification of the original algorithms. The calculated values of these entropies are not the same as the original ones, but the entropy trend of the analyzed time series determines equally reliably. Their main advantage is their speed, which is up to a thousand times higher. A scientific article describing their properties has been submitted to The Journal of Supercomputing and in present time it is waiting for the acceptance.

Early generation breeding trials are to be conducted in multiple environments where it may not be possible to replicate all the lines in each environment due to scarcity of resources. For such situations, partially replicated (p-Rep) designs have wide application potential as only a proportion of the test lines are replicated at each environment. A collection of several utility functions related to p-Rep designs have been developed. Here, the package contains six functions for a complete stepwise analytical study of these designs. Five functions pRep1(), pRep2(), pRep3(), pRep4() and pRep5(), are used to generate five new series of p-Rep designs and also compute average variance factors and canonical efficiency factors of generated designs. A fourth function NCEV() is used to generate incidence matrix (N), information matrix (C), canonical efficiency factor (E) and average variance factor (V). This function is general in nature and can be used for studying the characterization properties of any block design. A construction procedure for p-Rep designs was given by Williams et al.(2011) <doi:10.1002/bimj.201000102> which was tedious and time consuming. Here, in this package, five different methods have been given to generate p-Rep designs easily.

This package provides methods for analyzing and using quartets displayed on a collection of gene trees, primarily to make inferences about the species tree or network under the multi-species coalescent model. These include quartet hypothesis tests for the model, as developed by Mitchell et al. (2019) <doi:10.1214/19-EJS1576>, simplex plots of quartet concordance factors as presented by Allman et al. (2020) <doi:10.1101/2020.02.13.948083>, species tree inference methods based on quartet distances of Rhodes (2019) <doi:10.1109/TCBB.2019.2917204> and Yourdkhani and Rhodes (2019) <doi:10.1007/s11538-020-00773-4>, the NANUQ algorithm for inference of level-1 species networks of Allman et al. (2019) <doi:10.1186/s13015-019-0159-2>, the TINNIK algorithm for inference of the tree of blobs of an arbitrary network of Allman et al.(2022) <doi:10.1007/s00285-022-01838-9>, NANUQ+ routines for resolving multifurcations in the tree of blobs to cycles as in Rhodes et al.(2024) <doi:10.1186/s13015-025-00274-w>, and the ECToBlob algorithm for inference of a network with no anomalous quartets of Allman et al. (2026) (forthcoming). Software announcement by Rhodes et al. (2020) <doi:10.1093/bioinformatics/btaa868>.

This package provides tools for simulating the biophysical effects of vessel-strikes on whales. The aim is to support the evaluation of marine policies limiting ship speeds through regions in which whales reside. This is important because ship strikes are a major source of lethality for several whale species, including the critically endangered North Atlantic right whale. In this analysis, whales are modelled with a four-layer system comprising skin, blubber, sub-layer (muscle or organ) and bone. Reasonable values for the material properties of these layers, along with other factors such as whale surface area and mass, are provided for a variety of whale species. Similarly, key values are provided for several ship types. The collision is modelled according to Newtonian dynamics, with stresses and strains within the whale layers being simulated over time. The simulation results are analyzed in the context of whale-strike data, to develop a Lethality Index for the whale in the modelled collision. For the underlying science, see Kelley and other "Assessing the Lethality of Ship Strikes on Whales Using Simple Biophysical Models." (2021) <doi:10.1111/mms.12745>. For more on the R code, see Kelley "`whalestrike`: An R package for simulating ship strikes on whales" (2024) <doi:10.21105/joss.06473>.

In many agricultural, engineering, industrial, post-harvest and processing experiments, the number of factor level changes and hence the total number of changes is of serious concern as such experiments may consists of hard-to-change factors where it is physically very difficult to change levels of some factors or sometime such experiments may require normalization time to obtain adequate operating condition. For this reason, run orders that offer the minimum number of factor level changes and at the same time minimize the possible influence of systematic trend effects on the experimentation have been sought. Factorial designs with minimum changes in factors level may be preferred for such situations as these minimally changed run orders will minimize the cost of the experiments. This technique can be employed to any half replicate of two level factorial run order where the number of factors are greater than two. For method details see, Bhowmik, A., Varghese, E., Jaggi, S. and Varghese, C. (2017) <doi:10.1080/03610926.2016.1152490>. This package generates all possible minimally changed two-level half-fractional factorial designs for different experimental setups along with various statistical criteria to measure the performance of these designs through a user-friendly interface. It consist of the function minimal.2halfFFD() which launches the application interface.

Computation of dendrometric and structural parameters from forest inventory data. The objective is to provide a user-friendly R package for researchers, ecologists, foresters, statisticians, loggers and other persons who deal with forest inventory data. The package includes advanced distribution fitting capabilities with multiple estimation methods (Maximum Likelihood, Maximum Product Spacing with ties correction methods following Cheng & Amin (1983), and Method of Moments) for probability distributions commonly used in forestry. Visualization tools with confidence bands using delta method and parametric bootstrap are provided for three-parameter Weibull distribution fitting to diameter data. Useful conversion of angle value from degree to radian, conversion from angle to slope (in percentage) and their reciprocals as well as principal angle determination are also included. Position and dispersion parameters usually found in forest studies are implemented. The package contains Fibonacci series, its extensions and the Golden Number computation. Useful references are Arcadius Y. J. Akossou, Soufianou Arzouma, Eloi Y. Attakpa, Noël H. Fonton and Kouami Kokou (2013) <doi:10.3390/d5010099>, W. Bonou, R. Glele Kakaï, A.E. Assogbadjo, H.N. Fonton, B. Sinsin (2009) <doi:10.1016/j.foreco.2009.05.032>, R. C. H. Cheng and N. A. K. Amin (1983) <doi:10.1111/j.2517-6161.1983.tb01268.x>, and R. C. H. Cheng and M. A. Stephens (1989) <doi:10.1093/biomet/76.2.385>.

This package provides comprehensive methods to calculate posterior probabilities, posterior predictive probabilities, and Go/NoGo/Gray decision probabilities for quantitative decision-making under a Bayesian paradigm in clinical trials. The package supports both single and two-endpoint analyses for binary and continuous outcomes, with controlled, uncontrolled, and external designs. For single continuous endpoints, three calculation methods are available: numerical integration (NI), Monte Carlo simulation (MC), and Moment-Matching approximation (MM). For two continuous endpoints, a bivariate Normal-Inverse-Wishart conjugate model is implemented with MC and MM methods. For two binary endpoints, a Dirichlet-multinomial model is implemented. External designs incorporate historical data through power priors using exact conjugate representations (Normal-Inverse-Chi-squared for single continuous, Normal-Inverse-Wishart for two continuous, and Dirichlet for binary endpoints), enabling closed-form posterior computation without Markov chain Monte Carlo (MCMC) sampling. This approach significantly reduces computational burden while preserving complete Bayesian rigor. The package also provides grid-search functions to find optimal Go and NoGo thresholds that satisfy user-specified operating characteristic criteria for all supported endpoint types and study designs. S3 print() and plot() methods are provided for all decision probability classes, enabling formatted display and visualisation of Go/NoGo/Gray operating characteristics across treatment scenarios. See Kang, Yamaguchi, and Han (2026) <doi:10.1080/10543406.2026.2655410> for the methodological framework.

This package provides a function collection to extract metadata, sectioned text and study characteristics from scientific articles in NISO-JATS format. Articles in PDF format can be converted to NISO-JATS with the Content ExtRactor and MINEr ('CERMINE', <https://github.com/CeON/CERMINE>). For convenience, two functions bundle the extraction heuristics: JATSdecoder() converts NISO-JATS'-tagged XML files to a structured list with elements title, author, journal, history, DOI', abstract, sectioned text and reference list. study.character() extracts multiple study characteristics like number of included studies, statistical methods used, alpha error, power, statistical results, correction method for multiple testing, software used. The function get.stats() extracts all statistical results from text and recomputes p-values for many standard test statistics. It performs a consistency check of the reported with the recalculated p-values. An estimation of the involved sample size is performed based on textual reports within the abstract and the reported degrees of freedom within statistical results. In addition, the package contains some useful functions to process text (text2sentences(), text2num(), ngram(), strsplit2(), grep2()). See Böschen, I. (2021) <doi:10.1007/s11192-021-04162-z> Böschen, I. (2021) <doi:10.1038/s41598-021-98782-3>, Böschen, I. (2023) <doi:10.1038/s41598-022-27085-y>, and Böschen, I. (2024) <doi:10.48550/arXiv.2408.07948>.

Estimate the abundance of cell clones from the distribution of lengths of DNA fragments (as created by sonication, whence `sonicLength'). The algorithm in "Estimating abundances of retroviral insertion sites from DNA fragment length data" by Berry CC, Gillet NA, Melamed A, Gormley N, Bangham CR, Bushman FD. Bioinformatics; 2012 Mar 15;28(6):755-62 is implemented. The experimental setting and estimation details are described in detail there. Briefly, integration of new DNA in a host genome (due to retroviral infection or gene therapy) can be tracked using DNA sequencing, potentially allowing characterization of the abundance of individual cell clones bearing distinct integration sites. The locations of integration sites can be determined by fragmenting the host DNA (via sonication or fragmentase), breaking the newly integrated DNA at a known sequence, amplifying the fragments containing both host and integrated DNA, sequencing those amplicons, then mapping the host sequences to positions on the reference genome. The relative number of fragments containing a given position in the host genome estimates the relative abundance of cells hosting the corresponding integration site, but that number is not available and the count of amplicons per fragment varies widely. However, the expected number of distinct fragment lengths is a function of the abundance of cells hosting an integration site at a given position and a certain nuisance parameter. The algorithm implicitly estimates that function to estimate the relative abundance.

Allows for building drug response models using screening data between bulk RNA-Seq and a drug response metric and two additional tools for biomarker discovery that have been developed by the Huang Laboratory at University of Minnesota. There are 3 main functions within this package. (1) calcPhenotype() is used to build drug response models on RNA-Seq data and impute them on any other RNA-Seq dataset given to the model. (2) GLDS() is used to calculate the general level of drug sensitivity, which can improve biomarker discovery. (3) IDWAS() can take the results from calcPhenotype() and link the imputed response back to available genomic (mutation and CNV alterations) to identify biomarkers. Each of these functions comes from a paper from the Huang research laboratory. Below gives the relevant paper for each function. The package is described in Maeser et al. (2021) "oncoPredict: an R package for predicting in vivo or cancer patient drug response and biomarkers from cell line screening data" <doi:10.1093/bib/bbab260>. calcPhenotype() - Geeleher et al, Clinical drug response can be predicted using baseline gene expression levels and in vitro drug sensitivity in cell lines. GLDS() - Geeleher et al, Cancer biomarker discovery is improved by accounting for variability in general levels of drug sensitivity in pre-clinical models. IDWAS() - Geeleher et al, Discovering novel pharmacogenomic biomarkers by imputing drug response in cancer patients from large genomics studies.

Application of the Partitioning-Around-Medoids (PAM) clustering algorithm described in Schubert, E. and Rousseeuw, P.J.: "Fast and eager k-medoids clustering: O(k) runtime improvement of the PAM, CLARA, and CLARANS algorithms." Information Systems, vol. 101, p. 101804, (2021). <doi:10.1016/j.is.2021.101804>. It uses a binary format for storing and retrieval of matrices developed for the jmatrix package but the functionality of jmatrix is included here, so you do not need to install it. Also, it is used by package scellpam', so if you have installed it, you do not need to install this package. PAM can be applied to sets of data whose dissimilarity matrix can be very big. It has been tested with up to 100.000 points. It does this with the help of the code developed for other package, jmatrix', which allows the matrix not to be loaded in R memory (which would force it to be of double type) but it gets from disk, which allows using float (or even smaller data types). Moreover, the dissimilarity matrix is calculated in parallel if the computer has several cores so it can open many threads. The initial part of the PAM algorithm can be done with the BUILD or LAB algorithms; the BUILD algorithm has been implemented in parallel. The optimization phase implements the FastPAM1 algorithm, also in parallel. Finally, calculation of silhouette is available and also implemented in parallel.