SKESA WASM

SKESA (Strategic K-mer Extension for Scrupulous Assemblies) compiled to WebAssembly using Emscripten. Runs genome assembly entirely in the browser or Node.js with no server required.

Quick Start

Browser

Serve the repository and open demo/index.html:

python3 -m http.server 8000
# Visit http://localhost:8000/demo/

Upload paired FASTQ files (or gzipped FASTQ) and click "Assemble".

Node.js

import { readFileSync } from 'fs';
import { gunzipSync } from 'zlib';

const { default: createSKESA } = await import('./build/skesa.js');

const stdout = [];
const stderr = [];
const skesa = await createSKESA({
  wasmBinary: readFileSync('./build/skesa.wasm'),
  noInitialRun: true,
  print: (t) => stdout.push(t),
  printErr: (t) => stderr.push(t),
});

// Write reads to virtual filesystem
const r1 = gunzipSync(readFileSync('reads_R1.fastq.gz'));
const r2 = gunzipSync(readFileSync('reads_R2.fastq.gz'));
skesa.FS.writeFile('/r1.fastq', r1);
skesa.FS.writeFile('/r2.fastq', r2);

// Run assembly
skesa.callMain([
  '--reads', '/r1.fastq,/r2.fastq',
  '--cores', '1',
  '--min_contig', '200',
]);

// Output is in stdout array (FASTA format)
const fasta = stdout.join('\n');

Building

Requires Docker with the Emscripten SDK image:

./compile-docker.sh

Outputs build/skesa.js and build/skesa.wasm.

Test

node test/test.mjs

Assembles T7 phage reads and verifies the output (1 contig, ~40kb genome).

WASM-specific Patches

The SKESA source code required several fixes for correct operation on wasm32 (where size_t is 32-bit vs 64-bit on x86-64):

1. Read storage bit-packing (common_util.hpp): Cast to uint64_t before left-shifting nucleotide values by >= 32 bits. Without this, half of all nucleotides in the read storage were corrupted due to ptrdiff_t being 32-bit on wasm32.

2. Kmer count packing (KmerInit.hpp, counter.hpp): Changed count storage from size_t to uint64_t. SKESA packs strand info and branching data into the upper 32 bits of kmer counts using shifts (<< 48, << 32), which overflow on 32-bit size_t.

3. Thread forcing (skesa.cpp): Forces single-threaded operation (ncores=1) on Emscripten builds.

License

SKESA is a United States Government Work and is in the public domain. See src/LICENSE.