Blog

Which medical device trials in China need approval from the national regulator, which need only a filing, and what each route requires.

Clinical genome means two things: a sequencing test run and interpreted under lab standards, and ClinGen, the public resource that decides which genes and variants are real. Here is how both work and how to use them on your own data.

A working guide to going from raw sequencing reads to an annotated, filtered variant list you can reason about: file formats, strand conventions, QC thresholds, annotation commands, and the failure modes that produce confident nonsense.

An extended panel measures a few dozen analytes. How to choose them, how to space the draws, and how to store the results so you can compute on them.

A concrete stack for analyzing your own whole-genome data: alignment, variant calling, annotation, pharmacogenomics, and polygenic scores, with the flags, file sizes, and failure modes that matter.

A working pipeline for turning 30x whole-genome sequencing reads into an aligned, called, benchmarked, annotated and filtered variant set, plus the triage rules that keep the short list short.

A specific panel, drawn quarterly, with the analytic variation and pre-analytic rules that decide whether a change in a number is real. What to order, what to skip, and how to handle the data.

How Regulation (EU) 2017/745 governs clinical investigations of medical devices, which studies it covers and when authorisation or notification is needed.

Genotyping arrays, exomes, and whole genomes fail in different ways and in different places. Here is where the errors live, how to check your own data, and what accuracy numbers mean per-variant rather than per-test.

A practical setup for loading your own WGS alignments, variant calls, and RNA-seq coverage into IGV and igv.js, with the indexing, reference-build, and chromosome-naming details that usually break first.

A working pipeline for taking raw whole-genome sequencing reads of your own DNA through alignment, variant calling, annotation, and a queryable table you can hand to an analysis agent.

Whole exome sequencing covers roughly 1-2% of the genome at high depth. What the files contain, which tools we use, and where the coverage gaps are.

A whole-genome sequencing sample is usually 4–10 mL of EDTA blood or a saliva kit; what comes back is 80–120 Gbp of reads, a CRAM, and a VCF with roughly 4–5 million variants. Here is how to judge the sample, the files, and the costs.

How 21 CFR Part 862 special controls separate analytical performance, clinical performance and claims about patient management for in vitro diagnostics.

Pharmacogenetic testing predicts how fast you metabolize certain antidepressants. How to call the star alleles yourself, and where the calls fail.

A working pipeline from FASTQ to an annotated variant list, with the QC steps and the filters that make 4.5 million variants readable.

A genome reader is really three things: the sequencer that produces reads, the aligner and variant caller that turn reads into a genome, and the browser that lets you look at it. Here is what each step produces, which tools we would run, and where they fail.

A technical account of pharmacogenetic panels, whole-genome sequencing, and the files they produce: coverage requirements, star-allele calling, CYP2D6 copy number, and where prices land in 2026.

A step-by-step guide to going from FASTQ files to gene-level counts, cell-composition estimates, and longitudinal comparisons using your own transcriptome data.

A working pipeline from FASTQ to annotated VCF for one human genome, with DeepVariant and GATK calling and benchmarking against GIAB truth sets.