← back to logs
RESEARCHDEV

W-ASAP: Building Switzerland's Viral Wastewater Surveillance Platform

>
Cover Image for W-ASAP: Building Switzerland's Viral Wastewater Surveillance Platform

W-ASAP: Building Switzerland's Viral Wastewater Surveillance Platform

Infrastructure for National-Scale Genomic Analysis

Working Period: October 2024 - January 2026

On January 30, 2026, I left ETH Zurich's D-BSSE after 17 months as Software Engineer building W-ASAP (Wastewater Analysis & Sharing Platform) - the first of its kind read-level online tool for wastewater early viral detection.

Wastewater As Soon As Possible

Delivering on the promise of wastewater surveillance: early variant detection at national scale - the only financially sustainable path to population-wide viral monitoring.

The Challenge

Wastewater surveillance is a powerful tool for early detection of viral variants circulating in a population. Switzerland, like many countries, sequences viral fragments from wastewater treatment plants nationwide. But turning raw sequencing data into actionable insights for public health teams was a bottleneck.

The status quo:

  • Cluster-only access: Analysis required running complex workflows on ETH's Euler HPC cluster, accessible only to bioinformatics experts
  • Days of turnaround: Researchers had to request data analysis from a small group with cluster access and internal knowledge
  • Isolated data: Wastewater data couldn't be directly compared with clinical variant definitions without manual, time-consuming steps

The goal was to transform this into something accessible to virologists, epidemiologists, and public health teams - without requiring terminal expertise.

The Solution: Browser-Based Variant Exploration

W-ASAP bridges the gap between cluster-based terminal workflows and browser-based interactive exploration. It leverages existing infrastructure while making the data accessible:

W-ASAP Architecture

Data Flow

  1. V-Pipe (bioinformatics pipeline) processes raw sequencing data on ETH's Euler cluster
  2. sr2silo converts genomic alignment files for query engine ingestion
  3. Data submitted to Loculus (sequence database) via S3 bucket
  4. WisePulse (Ansible) ingests data into per-virus LAPIS-SILO (genomic query engine) databases
  5. LAPIS-SILO instances serve queries via REST API
  6. GenSpectrum (pathogen analysis platform) dashboards provide public web interfaces

The key innovation: first application of LAPIS-SILO to short read sequencing data. LAPIS-SILO, built by the GenSpectrum team, is a high-performance genomic query engine previously used for consensus-level data. Applying it to read-level wastewater data required careful engineering of the data pipeline and memory optimization.

What I Built

I took this project from concept to proof-of-concept to production. Three main components:

sr2silo - Python ETL pipeline converting genomic alignments for database ingestion. Published on Bioconda, runs daily on ETH's Euler cluster.

WisePulse - Ansible-based infrastructure as code for all W-ASAP services. Multi-virus pipeline with automatic rollback. Adding new viruses is now configuration, not engineering.

V-Pipe Scout (live demo) - Streamlit prototyping frontend for variant exploration. Served as testing ground before GenSpectrum integration.

Core Technology: LAPIS-SILO

github.com/GenSpectrum/LAPIS-SILO - Built by the GenSpectrum team (cEVO, ETH Zurich)

LAPIS-SILO is a high-performance genomic query engine written in C++. Under the hood, each column value gets a bitmap (one bit per row) indicating which rows contain that value. Queries are answered by bitwise AND/OR-ing the relevant bitmaps together, with Roaring Bitmap compression keeping it compact and cache-friendly.

W-ASAP represents its first application to short read sequencing data - a significant technical milestone that required:

  • 88% memory reduction (310GB → 37GB) - proving feasibility despite initial concerns
  • Multi-virus support: Path-based routing (/covid/, /rsva/) for separate databases
  • Overnight index updates with automatic rollback on failure

Achievements

  • 99.77% uptime (December 2025)
  • 88% memory reduction enabling cost-effective hosting on Hetzner
  • COVID and RSV-A live with daily updates
  • From days to minutes: Self-service exploration instead of waiting for expert intermediaries
  • Configuration over engineering: Adding new viruses requires only YAML config changes

Live Deployments

COVID Dashboard

genspectrum.org/swiss-wastewater/covid

RSV-A Dashboard

genspectrum.org/swiss-wastewater/rsv-a

LAPIS API

  • COVID: lapis.wasap.genspectrum.org/covid/
  • RSV-A: lapis.wasap.genspectrum.org/rsva/

Resources

ETH Zürich D-BSSE Feature

ETH D-BSSE Image Film Watch on YouTube - Representing Switzerland's national viral wastewater surveillance

Platform Walkthrough

youtu.be/kCUd-o1FbXg - Overview of the platform and analysis capabilities

Workshop Materials

github.com/gordonkoehn/wasap-workshop - Jupyter notebooks for hands-on exploration

API Documentation

Team

This work was a collaboration between two groups at ETH Zurich:

CBG (Computational Biology Group, D-BSSE):

  • Gordon J. Köhn (me) - Software Engineering, sr2silo, WisePulse, V-Pipe Scout
  • Ivan Topolsky - V-Pipe lead, overall technical ownership

cEVO (Computational Evolution, D-BSSE):

  • Chaoran Chen - GenSpectrum infrastructure
  • Felix Hennig - GenSpectrum dashboards, Loculus deployments, monitoring
  • Alexander Taepper - LAPIS-SILO optimization, data architecture

What's Next

The platform continues under CBG and cEVO. The roadmap includes:

  • RSV-B pipeline and dashboard (configs ready)
  • Influenza integration (pending metadata resolution)
  • Expanded COVID lookback window (currently 4 months)
  • Usage analytics for stakeholder reporting

The infrastructure I built at ETH Zurich transformed wastewater surveillance from an expert-only cluster workflow into a self-service browser tool. It's live, it's stable, and it's serving Switzerland's public health community.