Blog

48 posts · 한국어 글 (Korean posts)

AllProteomicsBioinformaticsNetwork BiologyBiomarker ResearchBioinformatics Tools
ProteomicsNov 14, 2026· 13 min read

Is 1% FDR Really 1%? I Hid 14,567 Arabidopsis Proteins in MaxQuant, FragPipe and DIA-NN Searches

I added an Arabidopsis proteome that cannot be in the sample to the search database of the same human, yeast and E. coli runs, and counted how often MaxQuant 2.8.1, FragPipe 24 and DIA-NN 2.3.2 reported it at 1% FDR. Precursors and peptides stayed well inside 1%. Protein groups did not always: the estimate was 2.09% for FragPipe, 1.33% for MaxQuant and 1.37% for DIA-NN without MBR. Match-between-runs transfers added little error.

#false discovery rate#FDR#entrapment#target-decoy
BioinformaticsNov 14, 2026· 8 min read

bcftools vs GATK HaplotypeCaller, Scored Against GIAB HG002: A Tie on SNPs, a Clear Gap on Indels, and Hard Filters That Cost Recall

We called germline variants on GIAB HG002 chromosome 20 with bcftools and GATK HaplotypeCaller and scored both against the GIAB v4.2.1 truth set with hap.py. On SNPs they tied (F1 0.992). On indels GATK was far ahead (0.992 vs 0.937), and most of bcftools' false indels had the right allele but the wrong genotype. GATK's recommended hard filters removed 580 true SNPs, mostly in low-mappability regions.

#variant calling#bcftools#GATK#HaplotypeCaller
BioinformaticsNov 13, 2026· 10 min read

Random Gene Signatures 'Predict' Survival in TCGA, and the 'Optimal' Cut-Point Makes Nearly Half of Useless Markers Significant

We scored 1,000 random 50-gene signatures against progression-free survival in four TCGA cohorts and simulated a marker with no link to survival. Picking the 'optimal' cut-point and reporting its log-rank p gave p < 0.05 for 44-49% of null markers. In kidney cancer, 57% of random signatures came out prognostic; with proliferation and stage in the model, that fell to 4-10% across the four cohorts.

#biomarker discovery#prognostic signature#survival analysis#Kaplan-Meier
BioinformaticsNov 9, 2026· 9 min read

Single-Cell Differential Expression Between Patients: Cell-Level Tests Flag Up to 24% of Genes When Nothing Differs. Pseudobulk vs FindMarkers on Real Donors

We split 8 real donors from the same condition into 4 vs 4 groups, 35 different ways, and ran cell-level tests (Seurat's default Wilcoxon and a Welch t-test, like Scanpy's default) and three pseudobulk tests. With no real difference, cell-level tests gave p < 0.05 for 9-24% of genes and on average up to 347 false discoveries per split. Pseudobulk stayed at or below 5%.

#single-cell RNA-seq#differential expression#pseudobulk#FindMarkers
BioinformaticsNov 6, 2026· 10 min read

Same Gene List, Different GO Enrichment: How Much the GO Release Changes the Result (2016 to 2025, Measured)

We ran the same gene lists through the same GO over-representation analysis with four Gene Ontology releases from 2016 to 2025. Planted answers and random lists behaved, but a real dexamethasone list kept only 26 of its significant terms, and 'response to glucocorticoid' was missed with the 2016 annotations.

#GO enrichment#Gene Ontology#enrichment analysis#pathway analysis
BioinformaticsNov 4, 2026· 10 min read

Correct Batch Effects, Then Test? With an Unbalanced Design, ComBat Turns a 5% False-Positive Target into 29%

Correcting batch effects with ComBat or removeBatchEffect and then testing the cleaned data looks harmless. We measured it on simulated data and on a real Bioconductor data set with fake labels: when groups are spread unevenly over batches, false positives rise from 5% to 28-29%, while putting batch in the limma model stays at 5%.

#batch effect#ComBat#removeBatchEffect#limma
ProteomicsOct 17, 2026· 17 min read

DIA-NN Quantification Modes Scored on Known Ratios: QuantUMS, Legacy MaxLFQ and Fragment-Level Re-Quantification, With and Without MBR

One set of DIA-NN 2.3.2 identifications, seven ways to turn it into protein quantities, scored against known ratios. Without MBR the MaxLFQ-type routes land within about 4 points of each other. With MBR, the strongly changed proteins MBR adds are compressed: QuantUMS handles them best, and re-quantifying them from raw fragments is much worse.

#DIA-NN#QuantUMS#MaxLFQ#directLFQ
ProteomicsOct 7, 2026· 10 min read

FragPipe vs MaxQuant, Measured: Same Six Runs, Same Known Answer (9 Minutes vs 62)

I ran FragPipe 24 (MSFragger 4.4.1 + IonQuant) and MaxQuant 2.8.1 on the same six Orbitrap runs of a three-species benchmark with known ratios. FragPipe finished in 8.9 minutes against 61.6, quantified 1.6–2.7× more proteins, and was more accurate on the proteins both tools quantified. What that does and does not mean, and the Linux setup traps.

#FragPipe#MaxQuant#MSFragger#IonQuant
ProteomicsOct 7, 2026· 12 min read

MaxQuant 2.8 proteinGroups.txt, Scored Against a Known Answer: LFQ vs iBAQ vs Intensity, and What Match-Between-Runs Adds

I ran MaxQuant 2.8.1 on six public Orbitrap runs where human, yeast and E. coli were mixed at known ratios, then scored every intensity column against the right answer. LFQ was far more accurate than raw Intensity but quantified fewer proteins, iBAQ ratios were identical to Intensity, MBR made already-quantified proteins more accurate, and the old Reverse filter silently returns zero rows in 2.8.

#MaxQuant#MaxQuant 2.8#proteinGroups.txt#LFQ intensity
BioinformaticsJun 5, 2026· 8 min read

Can Flux Balance Analysis Predict Antibiotic Synergy? 107,296 Simulations Say No — and What to Do Instead

A companion to two ESKAPE preprints: standard LP-based flux balance analysis structurally cannot detect synergy between essential gene pairs (107,296 simulations across 3 pathogens, zero synergy found). The fix is to stop using FBA as a synergy calculator and use it as a feature generator — partial-inhibition simulations plus ML to curate drug targets, fully open-source.

#flux balance analysis#antibiotic synergy#ESKAPE#genome-scale metabolic model
BioinformaticsJun 2, 2026· 6 min read

Can an LLM Run an RNA-seq Analysis on Its Own? Building ARIA, a Decision-Aware Transcriptome Framework

A companion to our preprint: reproducible pipelines run RNA-seq steps reliably, but the decisions between steps still need an expert. ARIA puts an LLM in that reasoning seat across 8 decision points — and on 4 public datasets it recovered paired designs, technical covariates, and known biology, with cross-method agreement r > 0.99.

#RNA-seq#transcriptomics#large language models#DESeq2
ProteomicsMay 27, 2026· 13 min read

limma vs DEqMS for Proteomics — When to Use Which (Measured at n=3 on Known Ratios)

Both limma and DEqMS provide moderated t-statistics for proteomics differential expression. limma was built for microarrays; DEqMS adds a proteomics-specific variance prior tied to peptide count. This guide scores both, plus the t-test, on a known-ratio benchmark at n=3 per group, with recommendations on when DEqMS's extra step is worth it and why a fold-change threshold matters.

#limma#DEqMS#proteomics statistics#moderated t-test
ProteomicsMay 23, 2026· 12 min read

BioMart Pig ↔ Mouse 1:1 Ortholog Mapping for Cross-Species Proteomics (R + Python Tutorial)

Step-by-step tutorial: download the Ensembl BioMart Release 111 Sus scrofa ↔ Mus musculus 1:1 ortholog table and join it to your per-species protein quantification in R (biomaRt) and Python (pandas). Why gene-symbol matching is fragile, how to handle one-to-many and many-to-many orthologs, and how to validate your mapping against UniProt.

#Ensembl BioMart#ortholog#Sus scrofa#Mus musculus
ProteomicsMay 23, 2026· 10 min read

Why You Must NOT Merge Species FASTA Databases in Cross-Species Proteomics (Shared Peptide Problem)

In cross-species LC-MS/MS proteomics (Porcine ECM vs Mouse Matrigel, human vs mouse cells, host-pathogen, etc.) merging species FASTA into one search database breaks protein quantification. This is the shared-peptide ambiguity problem. Here's why it happens, how to set up species-separated searches in MaxQuant and DIA-NN, and how to handle peptides that genuinely match both species.

#cross-species proteomics#MaxQuant#DIA-NN#FASTA database
ProteomicsMay 23, 2026· 11 min read

Reanalyzing PRIDE PXD023694 — Matrigel Nuclear Contaminants (EWSR1, RUVBL2) You'll Find in Real Data

A practical reanalysis of PRIDE dataset PXD023694 (Cho lab cross-species ECM) — what proteins actually come out of mouse Matrigel vs porcine ECM, the basement membrane signature (LAMB1/NID1/LAMC1/HSPG2) that replicates Park et al. 2026, and the contamination patterns (Matrigel nuclear proteins EWSR1/RUVBL2, porcine smooth muscle remnants TPM4/MYL9) that simulations miss but real data shows.

#PRIDE#PXD023694#Matrigel#EWSR1
ProteomicsMay 22, 2026· 14 min read

From DIA-NN Output to Paper Draft: A Complete AI-Assisted Proteomics Workflow (2026)

An honest, end-to-end guide to using AI across the full DIA-NN proteomics pipeline — from report.tsv quantification and downstream statistics through figure interpretation, Discussion writing, and a manuscript draft. Includes real prompts, where AI fails (with caught examples), verification checklists, and 2026 journal disclosure requirements.

#DIA-NN#proteomics#AI-assisted research#MaxLFQ
ProteomicsMay 19, 2026· 13 min read

Reproducing Park et al. 2026: Three Iterations of a Cross-Species ECM Proteomics Pipeline

A working record of reproducing the Park et al. 2026 cross-species extracellular matrix proteomics workflow (Porcine esophagus EEM vs Mouse Matrigel) — including three simulation iteration failures (circular logic, pseudocount artifact, gene-symbol vs Ensembl BioMart ortholog mapping) and independent PRIDE PXD023694 validation. Practical lessons for MaxQuant + riBAQ + DEP analysis.

#proteomics#cross-species#ECM#Matrigel
Biomarker ResearchMar 30, 2026· 27 min read

Biomarker Discovery and Validation: A Comprehensive Guide to the Workflow and Methods (2026)

Complete biomarker discovery and validation workflow — biomarker discovery methods (proteomics, metabolomics, genomics), study design, statistical pitfalls, regulatory pathway (FDA/CE-IVD/MFDS), and validation steps that separate publishable findings from clinically useful tests. Includes REMARK/TRIPOD checklists and real failure case studies.

#biomarker discovery#biomarker validation#biomarker discovery workflow#biomarker discovery methods
ProteomicsFeb 25, 2026· 14 min read

limma for Proteomics in R: A Complete Differential Expression Pipeline, Tested on MaxQuant and DIA-NN Output

Step-by-step limma differential expression for proteomics in R, from a MaxQuant or DIA-NN protein matrix to a volcano plot, with filtering, normalization, BH correction and three-group contrasts. Every block was run on real data with known ratios, which showed that normalizing DIA-NN output a second time and a log2FC cutoff of 1 can hide most true 2-fold changes.

#proteomics#differential expression#limma#R