Skip to content

Search Space

Jump to bottom
Charlie Bayne edited this page Mar 30, 2026 · 8 revisions

Search Space

Conduit supports multiple strategies for defining the protein search space. There is no single "best" method — the optimal strategy depends on your experimental design and what data you have available.

Available options:

Method Config value Input required
NCBI Taxonomy IDs ncbi_taxonomy_id input/ncbi_taxa_ids.txt
UniProt Proteome IDs uniprot_proteome_id input/proteome_ids.txt
Peptidotyping peptidotyping None (uses MS data)
MetaPhlAn metaphlan input/fastq_files/
MAGs MAGs input/MAG_files/

Recommendations

  • If MAGs are available and derived from your experimental system, these are often the best choice as they are most specific to your sample.
  • If the sample composition is known a priori (e.g., in vitro or gnotobiotic experiments), define the search space at the lowest practical taxonomic level — ideally using uniprot_proteome_id or ncbi_taxonomy_id.
  • If you have shotgun metagenomic data, MetaPhlAn performs well for taxonomic profiling.
  • If no sequence data are available, peptidotyping is a good option, leveraging MS data directly to infer taxa.

ncbi_taxonomy_id

Builds the protein database from UniProt proteomes matching the provided NCBI taxon IDs.

Input file: experiments/{experiment}/input/ncbi_taxa_ids.txt

Single-column, tab-delimited with header:

ncbi_taxa_id
83333
9606
10090

How proteomes are selected from NCBI taxonomy IDs

UniProt proteomes originate from genome assemblies submitted to NCBI. Because many labs may submit assemblies for the same organism, redundancy is common. UniProt performs clustering and selects representative proteomes to reduce redundancy (see UniProt Proteomes Help).

Conduit's selection logic:

  1. Check for a reference proteome — query the UniProt API for a reference proteome linked to the given taxonomy ID.
  2. Handle strain-level taxonomy — if the ID represents a strain and no reference proteome exists, move up to the species level and query again.
  3. Fallback to non-reference proteomes — if no reference proteome is found, query for other proteomes (redundant, representative, or excluded) and rank by curation score, BUSCO completeness, and protein count. Select the top-ranked entry.
  4. Resolve redundancy — if the chosen proteome is marked redundant, download the parent proteome instead.

Some NCBI taxonomy IDs have no suitable UniProt proteome and will be skipped.

uniprot_proteome_id

Builds the protein database directly from a user-supplied list of UniProt proteome IDs.

Input file: experiments/{experiment}/input/proteome_ids.txt

One proteome ID per line:

UP000000625
UP000005640
UP000000589

peptidotyping

Taxon-specific peptidotyping (TSP) defines the search space directly from your MS data in a two-pass approach:

  1. All UniProt proteins are digested in silico with trypsin via Unipept, and each peptide is assigned a lowest common ancestor (LCA).
  2. Peptides with LCAs at the species level or lower are retained, forming a species-specific peptide database.
  3. MS data are searched with DIA-NN using this peptide database.
  4. Peptides that pass detection thresholds are used to infer NCBI taxon IDs.
  5. Those taxon IDs are passed to the ncbi_taxonomy_id workflow to build the final database.

No additional input files required beyond the standard .raw files.

Key parameters

Parameter Default Description
presence_min_peptides 3 Minimum unique peptides to call a species present
presence_min_coverage 0 Minimum protein coverage to call a species present (0 = disabled)
min_taxon_db_peptides 10 Minimum species-specific peptides a taxon must have in the reference database

metaphlan

Runs MetaPhlAn on shotgun metagenomic FASTQ files to profile the microbial community, then builds a protein database from the detected taxa.

Input: FASTQ files placed in experiments/{experiment}/input/fastq_files/

MetaPhlAn uses clade-specific marker sequences to identify taxa present in the sample, producing conservative and low false-positive results. The output taxa list (filtered by abundance threshold) is passed to the ncbi_taxonomy_id workflow.

Key parameters

Parameter Default Description
relative_abundance_threshold 0.01 Minimum relative abundance (fraction) to include a taxon

MAGs

Uses metagenome-assembled genomes (MAGs) provided by the user. Bakta annotates the MAGs and the predicted proteins form the search space.

Input: MAG FASTA files placed in experiments/{experiment}/input/MAG_files/

This method uses a separate annotation pipeline that does not rely on UniProt, making it suitable for experimental-specific genomes and environments with limited UniProt coverage.

Key parameters

Parameter Default Description
bakta_db_type "light" Bakta database: "full" (~84 GB) or "light" (~2 GB)
bakta_db_dir resources/bakta/db/ Path to the Bakta database

Universal search space options

These apply to all methods:

Parameter Default Description
host_organism_ids [] NCBI taxon IDs to classify as "host" rather than "microbiome". E.g., human = 9606, mouse = 10090.
append_additional_proteome_id FALSE Append an extra UniProt proteome ID. E.g., human = UP000005640.
append_additional_ncbi_taxa_id FALSE Append an extra NCBI taxon ID.
exclude_proteome_id FALSE Exclude a specific UniProt proteome ID from the database.
exclude_ncbi_taxa_id FALSE Exclude a specific NCBI taxon ID from the database.

Clone this wiki locally