Run using a genomic surveillance configuration

In the previous tutorial, you learned how to analyze a small set of GISAID (“custom”) data in the context of a small set of reference data. For genomic surveillance applications, you will often focus your analysis on a set of data specific to your question of interest. For example, an analysis of SARS-CoV-2 circulation in a specific geographic area requires a focal set of sequences and metadata from that area.

In this tutorial, you will learn to define and analyze a focal set of data from a geographic division in the United States using a global genetic context. You will also learn how to define a genetic context that prioritizes sequences that are genetically similar to your focal set.


  1. Run using custom data. This tutorial introduces concepts expanded by the following tutorial.

  2. Register for a GISAID account, if you do not have one yet. However, registration may take a few days. Follow alternative data preparation methods in place of Curate data from GISAID, if you wish to continue the following tutorial in the meantime.


If you are not already there, change directory to the ncov directory:

cd ncov

and activate the nextstrain conda environment:

conda activate nextstrain

Curate data from GISAID

We will download a focal set of Idaho sequences from GISAID’s EpiCoV database.

  1. Navigate to GISAID, Login, and go to EpiCoV > Search.

    GISAID EpiCoV Search
  1. Filter to sequences that pass the following criteria:

    1. From Idaho, USA

    2. Collected within the last month (e.g. 2022-06-03 to 2022-07-01)

    3. Has a complete genome

    4. Has an exact collection date

    GISAID EpiCoV filter and select sequences

    Example of GISAID filters using collection date between 2022-03-01 and 2022-04-01


    If your selection has more than 200 sequences, adjust the minimum date until it has 200 sequences or less. This ensures the tutorial does not take too long to run.

  2. Select the topmost checkbox in the first column to select all sequences that match the filters.

  3. Select Download > Input for the Augur pipeline > Download.

  4. Download/move the .tar file into the ncov/data/ directory.

  5. Extract by opening the downloaded .tar file in your file explorer. It contains a folder prefixed with gisaid_auspice_input_hcov-19_ containing two files: one ending with .metadata.tsv and another with .sequences.fasta.

  6. Rename the files as idaho.metadata.tsv and idaho.sequences.fasta.

  7. Move the files up to the ncov/data/ directory.

  8. Delete the empty gisaid_auspice_input_hcov-19_-prefixed folder and the .tar file if it is still there.

Run the workflow

From within the ncov/ directory, run the ncov workflow using a pre-written --configfile:

nextstrain build . --cores all --configfile ncov-tutorial/genomic-surveillance.yaml

Break down the command

The workflow can take several minutes to run. While it is running, you can investigate the contents of genomic-surveillance.yaml (comments excluded):

  - name: reference_data
  - name: custom_data
    metadata: data/idaho.metadata.tsv
    sequences: data/idaho.sequences.fasta
  - name: background_data

  root: "Wuhan-Hu-1/2019"

    title: "Idaho-specific genomic surveillance build"
    subsampling_scheme: idaho_scheme
    auspice_config: ncov-tutorial/auspice-config-custom-data.json

      query: --query "(custom_data == 'yes')"
      max_sequences: 50
      query: --query "(custom_data != 'yes') & (country == 'USA')"
      max_sequences: 10
      group_by: division year month
        type: proximity
        focus: custom_sample
      query: --query "(custom_data != 'yes')"
      max_sequences: 10
        type: proximity
        focus: custom_sample

This configuration file is similar to the previous file. Differences are outlined below, broken down per configuration section.


  1. The file paths in the second input are changed to idaho.metadata.tsv and idaho.sequences.fasta.

  2. There is an additional input background_data for a regional North America dataset built by the Nextstrain team, for additional context.


The output dataset is renamed idaho, representative of the new custom data in the second input.

  1. The title is updated.

  2. There is a new entry subsampling_scheme: idaho_scheme. This is described in the following section.


This is a new section that provides a subsampling scheme idaho_scheme consisting of three subsamples. Without this, the output dataset would use all the provided data, which in this case is thousands of sequences that are often disproportionally representative of the underlying population.

  1. custom_sample

    • This selects at most 50 sequences from the custom_data input.

  2. usa_context

    • This selects at most 10 sequences from the USA from the background_data and reference_data inputs.

    • Sequences are subsampled evenly across all combinations of division, year, month, with sequences genetically similar to custom_sample prioritized over other sequences.

  3. global_context

    • This selects at most 10 sequences outside the USA from the background_data and reference_data inputs.

    • As with the usa_context above, this rule prioritizes sequences for the global context that are genetically similar to sequences in the custom_sample.

Visualize the results

Run this command to start the Auspice server, providing auspice/ as the directory containing output dataset files:

nextstrain view auspice/

Navigate to The resulting dataset should show the recent Idaho sequences against a backdrop of historical sequences:

Phylogenetic tree from the "genomic surveillance" tutorial as visualized in Auspice