khushi singh.in

Colophon

How this site is made

Static pages, a little JavaScript where it earns its place, open data, and illustrations drawn from scratch.

Principles

  • Truth before glow. Scientific claims carry evidence labels and references. Hypotheses are labelled as hypotheses.
  • No lab data. This site shows research context only. It never shows unpublished results, lab images or anyone’s private data.
  • Open data. Visuals are built from open sources (e.g. Protein Data Bank structures, FlyBase, PubMed) or drawn from scratch, with credits.
  • Fast and light. Pages are plain HTML. Heavy 3D models load only when you ask for them.
  • Private by design. No cookies, no trackers, no third-party fonts or embeds. The one exception is the AlphaFold explorer, which fetches live data from AlphaFold DB and UniProt (plus PDBe and the EBI Proteins API on request) only when you use it (details).
  • Readable by people and machines. Every page has structured data (schema.org) and a plain markdown copy (add .md to its address), and AI assistants get a map of the site at /llms.txt. Search engines and AI crawlers are welcome; the text is CC BY 4.0.

Built with

  • Astro (static site), Svelte (interactive islands) and Tailwind CSS
  • <model-viewer> and three.js for 3D and augmented reality
  • D3 for charts, drawn at build time so they need no JavaScript
  • Pagefind for private, static search
  • Live protein data from AlphaFold DB (including AlphaMissense) and UniProt, with experimental structures from PDBe and reported variants from the EBI Proteins API, analysed and drawn in your browser
  • The ESM-2 protein language model (Meta AI, MIT licence), run on your device with ONNX Runtime Web (MIT)
  • UTIF.js (MIT) for reading microscope TIFF images in the browser
  • The tools are written for this site and run on your device: the Image lab follows ImageJ’s definitions and exports ImageJ macros, the statistics are checked against R, and the primer Tm against Biopython
  • Typefaces: Fraunces, Atkinson Hyperlegible Next & Mono (Braille Institute), Mukta and Tiro Devanagari Hindi, all self-hosted

Data and image credits

Every visual credits its own sources on its page. The Image lab demo images come from the Broad Bioimage Benchmark Collection (BBBC022 and BBBC039, CC0), and the sequence demo is the GFP mRNA (ENA M62653). Molecular structures come from the RCSB Protein Data Bank (data CC0). Gene-expression data come from FlyBase (CC BY 4.0). Publication counts come from PubMed via NCBI E-utilities.

Use of AI

Parts of this site were drafted with AI assistance and reviewed by me; all scientific claims are checked against cited sources. No AI-generated images or data are used. Illustrations are drawn by hand in code or generated from real, cited data.

Colour and accessibility

Text colours meet WCAG 2.2 AA contrast in both themes. Charts use colour-blind-safe palettes (Okabe–Ito, viridis), and fluorescence-style art can be re-coloured for colour-blind readers. See the accessibility statement.