Blog

A practical pipeline for turning vendor raw files from a plasma proteomics run into an FDR-controlled protein quantification matrix you can model, with the tools, flags, and failure modes named.

A specific, opinionated toolchain for a single person's whole-genome, RNA-seq, proteomics, and CGM data: the tools we would run, the flags that matter, and the failure modes that silently corrupt results.

A technical guide to prescription-free continuous glucose monitors in the US — Stelo, Lingo, Libre Rio — with real accuracy numbers, cost per month, interference modes, and how to get the raw data out.

A step-by-step guide to processing personal plasma proteomics data, from raw mass spectrometry files or affinity panel exports through quantification, quality control, normalization, and longitudinal modeling in R.

A working guide to taking raw RNA-seq counts from your own samples through filtering, effective-length correction, TMM or median-of-ratios scaling, and variance stabilization, with the diagnostics that tell you whether the normalization held.

Fasting glucose above post-meal glucose is a common and mostly explainable pattern: overnight hepatic glucose output, dawn hormones, sensor and meter measurement error, and meal composition. Here is how to tell them apart with CGM data and lab numbers.

A practical walkthrough of working with raw sequencing reads: reading the format, checking quality, aligning to a reference, and knowing when the data is good enough to call variants.

Psychiatric pharmacogenetics is mostly two genes, CYP2C19 and CYP2D6, plus two HLA alleles. Here is what the evidence supports, how to call those genes from whole-genome data with PharmCAT and Cyrius, and where the commercial panels overreach.

A practical answer to RNA-seq depth: 30-40 million paired-end reads for whole-blood expression profiling, why the number moves with your question, and how to check that you got what you paid for.

A working guide to turning raw RNA sequencing reads from your own blood into quality-controlled gene and transcript quantifications, with the specific tools, flags, and reference files we would use, plus the failure modes that silently corrupt results.

A testosterone test is only interpretable if you control the timing, the assay, and the companion measurements. Here is the panel we would order, why free testosterone by calculation beats direct immunoassay, and how to read the numbers.

How 21 CFR Parts 803, 806 and 822 govern adverse event reporting, corrections and removals and ordered postmarket surveillance for US devices.

A technical comparison of genotyping arrays, exome sequencing and whole-genome sequencing, with the quality metrics to demand from each.

The national fees for registering a drug or medical device in China, when they fall due, and the exemption for small and micro enterprises.

A working guide to the files a sequencer produces and what to do with them: FASTQ through alignment, variant calling, annotation, and RNA quantification, with the commands, parameters, and failure modes that matter.

A pharmacogenetic test genotypes a small set of drug-metabolism genes and converts them into star-allele diplotypes and metabolizer phenotypes. Here is what that pipeline looks like, which genes are hard to call, and what the result does and does not tell you.

A survey of the labs selling pharmacogenomic testing, what their panels genotype, how strong the clinical evidence is, and how to reproduce (and extend) those calls from your own whole-genome data with PharmCAT, Cyrius, and Aldy.

A low red cell distribution width means your red blood cells are unusually uniform in size. How to read it alongside MCV, and when it is an artifact.

A decade-by-decade view of which blood measurements carry information, how often to repeat them, and why interpretation gets harder after 70. Written for people who want to keep their own longitudinal record.

A technical read of the comprehensive metabolic panel: the fourteen analytes, the derived quantities worth computing, and the common errors.