Blog
Technical articles on biomedical data science, machine learning pipelines, and reproducible research.
BPCells, LogNormalize and Integration: Design Decisions Behind a Single-Nucleus RNA-seq Pipeline in R
June 01, 2026 · 11 min read
Why a snRNA-seq pipeline on a 16 GB laptop is built on BPCells on-disk matrices, LogNormalize instead of SCTransform, and a side-by-side RPCA/Harmony integration comparison — the reasoning, not the boilerplate.
Read moreBulk RNA-seq with DESeq2: The Design Decisions That Make Differential Expression Trustworthy
June 01, 2026 · 9 min read
What actually decides whether a DESeq2 result holds up: the design formula, log-fold-change shrinkage, and a DESeq2-vs-edgeR cross-check — applied to 192 5xFAD Alzheimer mouse samples. The combined-tissue model returns a cross-validated microglial (disease-associated) signature: 418 DE genes in DESeq2, 465 in edgeR, 417 concordant.
Read moreFrom MaxQuant to Results: How I Structure an LFQ Proteomics Pipeline in R
May 28, 2026 · 14 min read
A walkthrough of the modular LFQ proteomics pipeline I use in practice — from MaxQuant ProteinGroups output through normalisation, mixed imputation, limma/eBayes differential abundance, and visualisation. Real code, real dataset, and the decisions that most tutorials quietly skip.
Read moreMissing Data Imputation in Label-Free Quantitative Proteomics: A Mixed Strategy Approach
May 07, 2026 · 12 min read
Missing values are unavoidable in label-free quantitative proteomics. Learn when and how to apply MNAR versus MAR imputation strategies using a robust mixed approach that classifies missingness patterns at the protein-condition level.
Read morelayout: default title: Blog permalink: /blog/ —
Blog
Technical articles on biomedical data science, machine learning pipelines, and reproducible research.
BPCells, LogNormalize and Integration: Design Decisions Behind a Single-Nucleus RNA-seq Pipeline in R
June 01, 2026 · 11 min read
Why a snRNA-seq pipeline on a 16 GB laptop is built on BPCells on-disk matrices, LogNormalize instead of SCTransform, and a side-by-side RPCA/Harmony integration comparison — the reasoning, not the boilerplate.
Read moreBulk RNA-seq with DESeq2: The Design Decisions That Make Differential Expression Trustworthy
June 01, 2026 · 9 min read
What actually decides whether a DESeq2 result holds up: the design formula, log-fold-change shrinkage, and a DESeq2-vs-edgeR cross-check — applied to 192 5xFAD Alzheimer mouse samples. The combined-tissue model returns a cross-validated microglial (disease-associated) signature: 418 DE genes in DESeq2, 465 in edgeR, 417 concordant.
Read moreFrom MaxQuant to Results: How I Structure an LFQ Proteomics Pipeline in R
May 28, 2026 · 14 min read
A walkthrough of the modular LFQ proteomics pipeline I use in practice — from MaxQuant ProteinGroups output through normalisation, mixed imputation, limma/eBayes differential abundance, and visualisation. Real code, real dataset, and the decisions that most tutorials quietly skip.
Read moreMissing Data Imputation in Label-Free Quantitative Proteomics: A Mixed Strategy Approach
May 07, 2026 · 12 min read
Missing values are unavoidable in label-free quantitative proteomics. Learn when and how to apply MNAR versus MAR imputation strategies using a robust mixed approach that classifies missingness patterns at the protein-condition level.
Read more