No packages match

gptr - Language Model Agents Inside the Live 'R' Session

Runs large language model agents inside the running 'R' session, so that models inspect and modify objects that are already in memory instead of re-running scripts from scratch. A single function, peter(), opens an interactive chat in the console or, given a prompt, runs the agent loop and returns a value that can be piped into further prompts. Supports the 'Anthropic', 'OpenAI' and 'Google Gemini' application programming interfaces, 'OpenAI'-compatible endpoints, the 'Claude Code' and 'Codex' command line tools, local 'Ollama' conversational models, and System One typed-decision models such as 'Jev', 'Clef' and 'Clef Flash' <https://typesafe.ai/blog/introducing-system-one-models-and-jev>, whose typed yes/no, choice and score answers can be used directly in if() and for() statements. Tools for reading, writing, editing and searching files and for evaluating 'R' code are implemented in 'R'; skills, extensions, Model Context Protocol ('MCP') servers, sub-agents and 'shiny' apps are supported. Sessions are recorded as 'R' scripts, 'R Markdown', 'Quarto' or 'Jupyter' documents that can be re-run. Writing files and evaluating model-generated code require the user's permission by default.

Last updated

6.53 score 2 stars 80 scripts 278 downloads

kdps - Kinship Decouple and Phenotype Selection (KDPS)

A phenotype-aware algorithm for resolving cryptic relatedness in genetic studies. It removes related individuals based on kinship or identity-by-descent (IBD) scores while prioritizing subjects with phenotypes of interest. This approach helps maximize the retention of informative subjects, particularly for rare or valuable traits, and improves statistical power in genetic and epidemiological studies. KDPS supports both categorical and quantitative phenotypes, composite scoring, and customizable pruning strategies using a fuzziness parameter. Benchmark results show improved phenotype retention and high computational efficiency on large-scale datasets like the UK Biobank. Methods used include Manichaikul et al. (2010) <doi:10.1093/bioinformatics/btq559> for kinship estimation, Purcell et al. (2007) <doi:10.1086/519795> for IBD estimation, and Bycroft et al. (2018) <doi:10.1038/s41586-018-0579-z> for UK Biobank data reference.

Last updated

3.70 score 219 downloads

bolt4jr - Interface for the 'Neo4j Bolt' Protocol

Querying, extracting, and processing large-scale network data from Neo4j databases using the 'Neo4j Bolt' <https://neo4j.com/docs/bolt/current/bolt/> protocol. This interface supports efficient data retrieval, batch processing for large datasets, and seamless conversion of query results into R data frames, making it ideal for bioinformatics, computational biology, and other graph-based applications.

Last updated

3.70 score 2 scripts 244 downloads

ggmugs - Visualization of Multiple Genome-Wide Association Study Summary Statistics

A 'grammar of graphics' approach for visualizing summary statistics from multiple Genome-wide Association Studies (GWAS). It offers geneticists, bioinformaticians, and researchers a powerful yet flexible tool for illustrating complex genetic associations using data from various GWAS datasets. The visualizations can be extensively customized, facilitating detailed comparative analysis across different genetic studies. Reference: Uffelmann, E. et al. (2021) <doi:10.1038/s43586-021-00056-9>.

Last updated

2.70 score 2 scripts 421 downloads