Claude Skills by HolobiomicsLab

Use when you have a GTF annotation file and need to systematically extract all local alternative splicing event coordinates (not transcript isoforms) for downstream PSI calculation or when you require a searchable index mapping events to contributing transcripts for inclusion and skipping forms.

Use when after running DESeq() and extracting base results with results(), apply shrinkage when you have differential expression estimates and want to reduce the variance of log fold change estimates while preserving signal.

Use when after obtaining shrunken or unshrunken log fold change estimates from DESeq2 results objects, particularly when comparing multiple shrinkage estimator types (apeglm, normal, ashr) or evaluating the effect of shrinkage on fold change estimates across genes with varying expression levels.

Use when when you have built a new tool implementation or major version and need to verify it produces equivalent results to a reference implementation on the same input data and index.

Use when you have a processed single-cell expression matrix (AnnData object) with pre-computed cluster assignments (e.g., leiden or louvain clusters in adata.obs) and want to discover which genes define each cluster's transcriptional identity.

Use when you need to verify that visualization functions produce graphically correct output that matches previously validated baseline images. Use it as part of automated testing workflows (e.

Use when you have generated gene-level count matrices via two methodologically distinct routes—e.

Use when your input is a single-cell gene expression matrix too large to fit in RAM, or you are working in a resource-constrained environment (e.g., shared compute cluster, laptop with limited memory). You have constructed or loaded an AnnData object with X as a dask.array.

Use when you have applied two or more competing analysis workflows to the same RNA-seq or microarray dataset and need to assess whether they yield consistent or divergent differential expression results. Typical triggers: (1) comparing a new normalization method (e.

Use when you have raw or normalized microarray expression matrices stored in public repositories (GEO accessions) along with sample metadata, and you need to load both into R as structured objects (expression matrix and phenotype data frame) to construct a design matrix and fit linear models for.

Use when when you have raw or minimally processed microarray expression data (e.g., from GEO) with intensity values that exhibit sample-to-sample distributional differences and variance heterogeneity.

Use when you have a fitted linear model (lmFit object) from microarray, RNA-seq, qPCR, or proteomics data and need to test for differential expression across genes while maintaining statistical stability despite having few biological replicates or arrays.

Use when when you have ranked gene statistics and gene set collections, and your analysis requires P-value discrimination below a fixed lower bound (e.g., distinguishing between pathways at p < 1e-10).

Use when you have a normalized gene expression matrix (bulk RNA-seq or microarray) from a time-course or multi-condition experiment and need to quantify whether known gene sets (pathways, functional modules) show statistically significant coordinated expression changes.

Use when after running fgsea() on a preranked gene list and obtaining a results table with raw pval, padj, ES, NES, and size columns, use this skill to: (1) subset results to top enriched pathways (e.g., top 10 upregulated ES > 0 and top 10 downregulated ES < 0 by adjusted p-value);

Use when when analyzing spatial molecular data (e.g., from tissue sections or microscopy) where you need to build k-nearest neighbor graphs on high-dimensional coordinate arrays (n_obs × n_dims) and exact nearest neighbor computation is too slow or memory-intensive;

Use when when you have spatial transcriptomics or imaging data (e.

Use when you have long-read RNA-seq samples quantified by oarfish (output as quant.gz files) and need to extract transcript-level or gene-level abundance, count, and length matrices for downstream statistical analysis.

Use when after using edgeR::DGEListFromTximport with divide=TRUE on tximport output containing Gibbs sample or bootstrap replicates.

Use when after running DESeq() to fit negative binomial GLMs and obtaining raw p-values from results(dds), when you need to reduce false positives from multiple testing across thousands of genes while maximizing statistical power.

Use when you have clustered single-cell RNA-seq data (via Leiden, Louvain, or equivalent) and a k-nearest neighbor graph computed in PCA space, and you want to abstract cell-level connectivity into cluster-level connectivity to infer developmental trajectories, lineage relationships, or.

Use when you have paired-end RNA-seq reads (FASTQ) and a reference transcriptome (FASTA), and you need to estimate transcript-level expression (NumReads and TPM).

Use when when you have run gene set enrichment analysis (e.

Use when you have a multi-cluster single-cell RNA-seq dataset with cell-type annotations and you want to test whether known biological pathways (e.g., KEGG or MSigDB gene sets) show significantly elevated or differential enrichment across cell types.

Use when you have a preranked gene list (e.g., genes sorted by log2 fold-change, t-statistic, or other continuous metric) and a collection of gene sets or biological pathways, and you need to determine which pathways are significantly over-represented among highly-ranked genes (positive enrichment).

Use when you have normalized gene expression data (log-transformed, quantile-normalized) from a time-course or case-control experiment, a ranked gene statistic (e.g., mean expression, differential expression score), and a collection of curated gene sets (e.

Use when when comparing transcript quantification outputs (NumReads counts, abundance estimates) from two mapper implementations (e.

Use when you have quantification results from two or more independent implementations, versions, or variants of the same analysis tool (e.g. C++ salmon 1.11.

Use when comparing mapping outputs from two different salmon versions or implementations (e.g., C++ 1.12.0 vs. Rust 2.0) to determine if observed differences in total mapped read counts are due to true algorithmic bugs, parameter defaults, or index format changes.

Use when you have raw RNA-seq read counts and a set of normalization factors (e.g., TMM-computed library size scales from edgeR's calcNormFactors), and you plan to fit a linear model to detect differential expression.

Use when your input is a normalized, centered gene expression matrix with many genes (e.g., 12,000+) and you need to validate whether reducing to a smaller number of principal components (e.g., 10) preserves pathway enrichment statistics (pathway scores and p-values from geseca or fgsea).

Use when you have a fitted linear model (lmFit object) from microarray or RNA-seq count data and need to compute differential expression statistics, especially when the number of biological replicates is small (fewer than ~5–10 arrays/samples per group) and you want to avoid inflated variance.

Use when you have transcript-level abundance estimates (from Salmon, kallisto, or similar) and a set of defined alternative splicing events (in ioe or ioi format), and you need to compute event-level or isoform-level inclusion ratios (PSI) to quantify splicing patterns across multiple samples or.

Use when you have generated PSI matrices for alternative splicing events or transcripts across two or more biological conditions using SUPPA's psiPerEvent or psiPerIsoform subcommand, and you need to align and standardize these matrices with corresponding transcript expression quantification files.

Use when after forking and cloning a repository (e.g., scverse/scanpy) to verify that the development environment is correctly configured, or after implementing a feature or bugfix to ensure no regressions were introduced.

Use when after executing a spatial analysis function (e.g., squidpy.gr.sepal) that modifies or augments a data object, verify that the expected output fields exist with correct names, data types, and array dimensions before proceeding to interpretation or visualization.

Use when you have transcript-level quantification output files (e.g., quant.sf from salmon, abundance.

Use when when comparing mapping performance between two mapper implementations (e.g., C++ salmon vs. Rust salmon), validating that a bug fix or algorithmic change did not degrade sensitivity, or establishing baseline quantification quality on benchmark datasets with known characteristics.

Use when after computing gene set enrichment scores (e.g., via GESECA on reverse PCA feature loadings) on a single-cell or bulk dataset with an existing dimensionality reduction (tSNE, UMAP, PCA).

Use when when evaluating a computational method described in a peer-reviewed article, particularly when the work references a public repository or provides supplementary code.

Use when after running DESeq() on a DESeqDataSet and obtaining initial results via results(), use this skill when you need to (1) extract base results tables for specific contrasts (e.

Use when you have transcript-level quantification output files (quant.sf, quant.gz) from salmon, kallisto, sailfish, or oarfish and need to produce gene-level count matrices, abundance matrices, and length-based offsets for differential expression analysis in DESeq2, edgeR, or limma-voom.

Use when you have transcript-level abundance estimates from salmon, sailfish, or kallisto quantification and need gene-level count matrices for differential expression analysis.

Use when apply TMM normalization when you have raw RNA-seq read counts from multiple samples and suspect differences in library composition (e.g., one sample over-represents a highly-expressed gene or transcript class relative to others).

Use when you have transcript-level abundance estimates and inferred counts from salmon/kallisto/Sailfish (with or without Gibbs/bootstrap replicates) that must be summarized to gene or transcript level, and you plan to use count-based differential expression tools (edgeR, DESeq2, limma-voom).

Use when you have RNA-seq read count data (from alignment tools, transcript quantification, or feature counting) organized in a count matrix with samples as columns and genes as rows, paired with sample metadata (condition, batch, treatment), and you want to test which genes show statistically.

Use when before constructing a DESeqDataSet from count data or tximport output, when you have raw RNA-seq samples that need to be annotated with experimental conditions, treatment groups, batch effects, or other covariates.

Use when you have paired-end or single-end RNA-seq reads (FASTQ format) and a reference transcriptome (FASTA), and you need to estimate transcript-level abundances (NumReads and TPM per transcript) rather than gene-level counts.

Use when you have quantification output from two versions or implementations of the same tool (e.g., C++ vs. Rust rewrites, pre- and post-bugfix releases) and need to verify that they produce equivalent results before deploying the new version.

Use when when comparing two implementations of the same RNA-seq mapping algorithm on identical reference indices and read sets, if per-read mapping agreement is <99.8% or the overall mapping rate differs by >0.