Claude Skills by lilinji

Builds and validates predictive time-to-event models on clinical and omics data with penalized Cox, random survival forests, gradient-boosted and deep survival models, and prediction-grade evaluation (Uno's C, time-dependent AUC, integrated Brier, calibration, competing risks). Use when building an individualized risk predictor or prognostic omics signature, choosing a survival model, or evaluating one beyond the C-index. For Kaplan-Meier, log-rank, and classical Cox hazard-ratio inference in...

Decision-grade statistical analysis for metabolomics intensity tables. Covers transformation and scaling (Pareto vs unit-variance as a hidden hypothesis), unsupervised structure (PCA/HCA for QC), permutation-validated PLS-DA/OPLS-DA (R2 vs Q2, double CV, VIP as heuristic), univariate testing (Welch/Mann-Whitney/ANOVA/LMM with covariate adjustment), and dependence-aware multiple testing. Use when testing which metabolites differ, building or validating a discriminant model, choosing a scaling,...

Performs probe filtering and sample-level QC on Illumina Infinium methylation arrays (450K / EPIC / EPICv2) to decide which probes and samples to trust. Drops detection-p-failed and low-bead-count probes, removes cross-reactive/non-specific probes (Chen 2013 / Pidsley 2016 lists via maxprobes), excludes SNP-overlapping probes with dropLociWithSnps, and handles sex-chromosome probes. Collapses EPICv2 replicate probes with betasCollapseToPfx and harmonizes across array versions (EPICv2 hg38 vs ...

Computes DNA methylation age (DNAm age) and pace of aging by applying frozen elastic-net epigenetic clocks to a clean beta matrix with methylclock, dnaMethyAge, or methylCIPHER. Covers the clock menu by question (chronological Horvath/Hannum/skin&blood; health-mortality PhenoAge/GrimAge; DunedinPACE pace; pediatric/gestational; mitotic epiTOC), age acceleration (EAA/IEAA/EEAA) as the real endpoint, the principal-component (PC) clock fix for the per-CpG reliability crisis, and EPICv2 clock-CpG...

Imports Bismark coverage or cytosine-report files into the methylKit object model, then runs the import-to-results spine - filterByCoverage, normalizeCoverage, unite/destrand, calculateDiffMeth, getMethylDiff - for both per-CpG (DMC) and fixed-tile (DMR) differential methylation, plus tileMethylCounts, PCA/correlation/clustering QC, and assocComp/removeComp batch handling. Covers the silent default traps that shape the false-positive rate: overdispersion='none' does no correction while 'MN' f...

Performs ML-based protein-ligand pose prediction and scoring using DiffDock-L (diffusion-based), Boltz-1 / Boltz-2 (foundation model with affinity), Chai-1, AlphaFold3 ligand, EquiBind, TANKBind, NeuralPLexer, and hybrid workflows (DiffDock pose + GNINA rescore + PoseBusters QC). Explicit handling of when ML beats classical docking, when classical beats ML, the PB-invalid pose problem, and rescoring as the standard production hybrid. Use when modern docking is needed: foundation-model ligand-...

Single-variant common-variant GWAS with plink2 --glm (linear/logistic, Firth) and the linear mixed models GEMMA, BOLT-LMM, SAIGE, regenie (SPA). A GWAS statistic is valid only when genotype is independent of unmodeled phenotype drivers after the chosen covariates and random effects, so the engine follows sample structure and case:control imbalance, not taste: PC covariates absorb continuous ancestry but cannot remove relatedness (a covariance structure needing an LMM), genomic inflation above...

Manages PLINK genotype filesets - format conversion (VCF, BED/BIM/FAM, PED/MAP, pgen/pvar/psam) and sample/variant QC (missingness, MAF, HWE, sex check, heterozygosity, KING relatedness) with PLINK 1.9 and 2.0. PLINK rewrites allele bookkeeping: PLINK 1.x A1 defaults to the minor allele and is recomputed every load, silently flipping effect-allele meaning unless --keep-allele-order, while PLINK 2.0 tracks explicit REF/ALT. QC order matters (variant before sample missingness), HWE is controls-...

Infers and describes population structure with PCA (plink2 --pca, smartpca/EIGENSOFT, FlashPCA2), model-based clustering (ADMIXTURE, fastSTRUCTURE), FST estimators (Weir-Cockerham vs Hudson), and f-statistics (f3/f4/D via AdmixTools/admixr), plus Python plotting of PCs and Q barplots. Every output is a model-conditioned description of variance, not truth: PCs conflate ancestry with LD/inversions/relatedness/batch, ADMIXTURE Q-values are panel- and K-dependent artifacts, and CV-minimum K is a ...

In-memory Python population genetics with scikit-allel - GenotypeArray/HaplotypeArray/AlleleCountsArray, diversity (pi, theta, Tajima's D), SFS, FST (Weir-Cockerham, Hudson, Patterson), f3/D admixture stats, LD pruning, PCA, and selection scans (iHS, XP-EHH, nSL, Garud H). Nearly every statistic is a ratio or density with one silent denominator bug in two faces: omit is_accessible= and per-base pi/theta divide by total span not accessible bp (deflated 2-5x); average per-SNP FST instead of sum...

Test whether cell-type proportions or composition changed between conditions in single-cell data using Milo (miloR), scCODA, sccomp, and propeller. Use when comparing cell-type proportions / composition between conditions, asking which populations expanded or contracted with treatment or disease, running neighborhood-level (cluster-free) abundance testing, or guarding against compositional shifts that masquerade as differential expression.

Detect and remove doublets (two or more cells in one droplet) from single-cell RNA-seq using scDblFinder (R), Scrublet (Python), and DoubletFinder (R). Use when flagging artificial intermediate populations before clustering, setting the expected doublet rate from recovered-cell counts, running detection per sample before integration, choosing between simulate-and-score methods, or interpreting a non-bimodal score histogram.

Align V(D)J reads and assemble TCR/BCR clonotypes with MiXCR, driven by a chemistry-matched preset. Use when choosing/auditing the preset for a library (5'RACE/template-switch vs multiplex-primer amplicon -> rigid vs floating boundaries; RNA vs gDNA -> --rna/--dna; bulk vs 10x single-cell; UMI vs no-UMI -> tag pattern and barcode collapse; kit presets Takara/NEBNext/QIAseq/BD/MiLaboratory); assembling clonotypes by CDR3 vs VDJRegion; setting the reads-vs-UMI-vs-cell quantitation denominator; ...

Models continuous temporal trajectories from BULK or time-resolved omics where the x-axis is measured experimental time: penalized GAMs (mgcv) for smooth trends and changepoint detection (segmented, ruptures) for abrupt regime shifts. Use when deciding between a smooth GAM and a changepoint model; choosing the GAM distribution (nb() plus a library-size offset for raw counts vs Gaussian on vst/log-CPM); setting the basis-dimension ceiling k below the number of timepoints and letting REML pick ...

Estimates tumor fraction (the genome-wide proportion of cfDNA molecules that are tumor-derived, the cfDNA analogue of bulk-tumor purity) from shallow whole-genome sequencing with ichorCNA, an HMM over 1 Mb bins that jointly EM-estimates tumor fraction, ploidy, and subclonal prevalence over a normal/ploidy grid. Encodes the load-bearing reframes: tumor fraction is the quantity that travels across assays and is NOT mutation VAF (clonal-het VAF approximately TF/2), CNA-based estimation has a har...

Orchestrates the copy-number pipeline from BAM to segmented, integer-called, annotated CNVs, forking on germline-vs-somatic - CNVkit (somatic exome/panel: coverage -> assay-matched reference/PoN -> fix -> segment -> purity/ploidy-aware call), GATK gCNV (germline rare-CNV cohort), and allele-specific callers (ASCAT/FACETS/PURPLE) for purity/ploidy. Use when committing the build + target/access BED + PoN once (assay-matched), building the reference from normals BEFORE segmenting, fitting purity...

Orchestrates genomic-epidemiology outbreak investigation from pathogen isolates to transmission networks, forking bacterial (snippy -> Gubbins recombination-masking -> IQ-TREE -> TreeTime -> TransPhylo) vs viral (Nextstrain/augur), with parallel MLST typing (cgMLST delegated to epidemiological-genomics/pathogen-typing) and AMR surveillance. Use when committing ONE reference genome for SNP calling (every isolate and distance inherits its coordinates), applying MANDATORY Gubbins recombination-m...

Orchestrates the end-to-end bulk RNA-seq differential-expression pipeline from FASTQ to an annotated DE gene table, chaining fastp QC/trim, Salmon (decoy-aware) or STAR+featureCounts quantification, tximport gene-level collapse, DESeq2/edgeR/limma-voom testing, apeglm shrinkage, and VST-based visualization. Use when committing the reference release and gene-ID namespace once for the whole run, sequencing steps in the defensible order (tximport before DE, raw counts into the model, VST only fo...

Search bioRxiv biology preprints with natural language queries. Semantic search powered by Valyu.

Load when removing batch effects from a multi-cohort bulk RNA-seq dataset using ComBat (R

Load when discovering gene co-expression modules and hub genes in a bulk RNA-seq cohort via

Load when comparing gene expression between two conditions in bulk RNA-seq count data. Skip

Load when estimating cell-type proportions in bulk RNA-seq samples from a single-cell or

Load when running pathway / GO term enrichment on a bulk RNA-seq DE result list. Skip when

Load when converting gene identifiers between Ensembl, Entrez, and HGNC symbol in a bulk

Load when querying STRING for the protein-protein interaction subgraph induced by a bulk

Load when checking a bulk RNA-seq count matrix for library-size outliers, gene detection

Load when summarising STAR / HISAT2 / Salmon alignment-rate logs in bulk RNA-seq. Skip when

Load when checking raw FASTQ quality (Phred / GC / adapter / Q20-Q30) before alignment in

Load when summarising rMATS / SUPPA2 alternative-splicing output and identifying significant

Load when stratifying patients by gene expression and testing for survival differences (Kaplan-Meier

Load when placing bulk RNA-seq samples on a single-cell reference's pseudotime axis (NNLS

Load when you want a verified multi-method consensus over spatial tissue domains on a preprocessed

Load when biologically interpreting a finished verified consensus run (consensus-domains

Load when computing alignment QC metrics (mapping rate, MAPQ distribution, insert size, duplicate

Load when computing genome-assembly QC metrics — N50/N90, L50/L90, total length, contig count,

Load when calling CNV segments via CBS-style segmentation on a bin-level log2-ratio CSV from

Load when summarising a peak file (BED / narrowPeak) from ATAC-seq / ChIP-seq / CUT&Tag —

Load when summarising a phased VCF (output of WhatsHap / SHAPEIT5 / Eagle2) — phased fraction

Load when running pre-alignment FASTQ quality control — Phred quality scores, Q20/Q30 rates,

Load when summarising structural variants from an SV VCF (DEL / DUP / INV / TRA) — BND-notation

Load when summarising functional impact of an annotated variant CSV — per-IMPACT counts (HIGH

Load when summarising small variants (SNVs / indels) from a VCF or computing demo-pattern

Load when summarising / filtering a VCF — variant classification (SNP / MNP / INS / DEL /

Use Hermes Tweet and Xquik for X/Twitter agent workflows. Plan social listening, account and follower analysis, post research, monitors, webhook alerts, REST API calls, MCP client setup, and safe tweet actions.

Load when extracting GEO accessions, dataset metadata, and downloadable references from a

Extract medical entities (symptoms, medications, lab values, diagnoses) from patient messages.

Query 14+ biomedical databases for drug repurposing, target discovery, clinical trials, and literature research. Access ChEMBL, PubMed, ClinicalTrials.gov, OpenTargets, OpenFDA, OMIM, Reactome, KEGG, UniProt, and more through a unified MCP endpoint. Use when researching disease targets, finding approved/investigational drugs, searching clinical evidence, discovering genetic associations, or analyzing compound bioactivity data.

Ensure deterministic and safe memory usage: static allocation, pools, stack sizing, MPU usage, and corruption detection for medical device firmware.

Load when annotating LC-MS features against a built-in 15-metabolite HMDB demo dictionary