pySec2Pri

Create and use mapping files for secondary (retired/withdrawn) biological database identifiers and labels to primary (current) identifiers and labels.

Outputs mappings in SSSOM format by default. Subjects are secondary, objects are primary.

Installation

uv pip install pysec2pri

Or install from source:

uv pip install git+https://github.com/jmillanacosta/pysec2pri.git

Quick Start

Generating mapping sets

Most sources have two commands. ids maps retired identifiers to current ones, labels maps old labels to current ones:

pysec2pri hgnc ids
pysec2pri hgnc labels

Run pysec2pri --help to see every source, and pysec2pri <source> ids --help for one source's options. Input files are downloaded unless you pass them:

pysec2pri hgnc ids --withdrawn withdrawn.txt --complete hgnc_complete_set.txt

In Python there are two functions, one per kind:

from pysec2pri import generate_ids, generate_labels, sources

sources()          # every source
sources("labels")  # sources with labels

hgnc = generate_ids("hgnc")
chebi = generate_labels("chebi", subset="3star", version="350")
ensembl = generate_ids("ensembl", version="115", species="9606")

Both return an SSSOM MappingSet: IdMappingSet or LabelMappingSet. Subjects are secondary, objects are primary. Options that a source does not have are ignored, so species on HGNC does nothing.

The default output is SSSOM TSV.

Updating IDs and labels

Use a mapping set to update your own data. Labels:

from pysec2pri import generate_labels, resolve_labels
chebi = generate_labels("chebi")
resolve_labels(["Glucose", "ATP", "Guanine"], chebi)

Identifiers in a dataframe:

from pysec2pri import generate_ids, update_ids
ensembl = generate_ids("ensembl", version="115", species="9606")
df_with_new_column = update_ids(mapping_set=ensembl, ids=df, at="Ensembl_id")  # `at` is the column name

Or from the command line, given a TSV file gene_ex.tsv:

gene	data
HGNC:131	3.5

Resolve the gene column to primary HGNC IDs (a new _primary column is added):

pysec2pri update-ids gene_ex.tsv hgnc --at gene -o gene_ex_primary.tsv
# gene        data    gene_primary
# HGNC:131    3.5     HGNC:145

The same pattern works for labels with update-labels, and multiple columns can be resolved by repeating --at:

pysec2pri update-ids data.tsv hgnc --at gene_id --at related_gene_id

To skip regenerating the mapping set, pass a pre-built mapping file:

pysec2pri hgnc ids  # outputs hgnc_{version}_sssom.tsv
pysec2pri update-ids gene_ex.tsv hgnc --at gene --mapping hgnc_{version}_sssom.tsv

In Python, load_mapping reads a written SSSOM file back in, so you can generate once and reuse the file:

from pysec2pri import load_mapping, update_ids
hgnc = load_mapping("hgnc_115_sssom.tsv")
df_with_new_column = update_ids(mapping_set=hgnc, ids=df, at="gene")

Use load_label_mapping for a label mapping set.

Ambiguous mappings (where a deprecated ID or label serves as a recommended for another entity) are not resolved, but flagged for users to solve them manually. If the input file has a column of known aliases or synonyms for each row, pass it as a hint to resolve ambiguous names automatically:

pysec2pri update-ids data.tsv hgnc --at gene_id --synonyms gene_aliases
# Pairs gene_aliases hints with gene_id; repeat --at X--synonyms Y for more columns.

A subset with ambiguous mappings only can be generated like:

pysec2pri ambiguous hgnc-labels

Mapping types

Every row is one secondary (subject) and one primary (object). Which predicate joins them says what happened to it.

IDs

A retired ID either has a replacement or does not:

flowchart LR
    S["subject_id (retired)"]
    P["object_id (current)"]
    N["sssom:NoTermFound"]
    S -->|"IAO:0100001 (term replaced by)"| P
    S -->|"oboInOwl:consider (no replacement)"| N

mapping_cardinality says how the two sides line up: 1:1, n:1 when several retired IDs were merged into one, 1:0 for a withdrawal with no replacement.

Labels

One label mapping set holds both of a source's label changes, told apart by predicate:

flowchart LR
    PREV["subject_label (previous symbol)"]
    ALIAS["subject_label (alias / synonym)"]
    CUR["object_label / object_id (current)"]
    PREV -->|"IAO:0100001 (term replaced by)"| CUR
    ALIAS -->|"oboInOwl:hasExactSynonym"| CUR

A previous symbol is one the entity used to have. An alias is another name it still goes by. Only the first is a rename; the second is what the resolver uses as evidence below.

Looking further back with --consolidate

--consolidate reads all of the source's past releases, finds mappings the current release no longer mentions, and gives every mapping the release it first appeared in:

pysec2pri hgnc ids --consolidate -o hgnc.sssom.tsv

from pysec2pri import generate_ids, supports_consolidate
supports_consolidate("hgnc", "ids")
generate_ids("hgnc", consolidate=True)

Ambiguity

A value is ambiguous when it is retired in one row and current in another. This is not resolved: the row is flagged and left alone.

flowchart LR
    C["C (retired)"] -->|term replaced by| A["A (retired, and current for C)"]
    A -->|term replaced by| B["B (current)"]

The same holds for labels: a symbol can be a subject_label (someone's old name) and an object_label (someone else's current name).

Resolving ambiguity with alias/synonym hints

When a name is ambiguous, alias mappings are used as evidence. For each candidate interpretation the resolver checks whether any user-supplied hint matches a known alias of that candidate's primary entity. A hit on the secondary candidate's target confirms the name is being used as a previous name; a hit on the primary candidate's aliases confirms it is already current.

flowchart TD
    Name["ambiguous name"]
    Hint["Alias hint"]
    Check{"Hint matches alias of…"}
    SecPath["Replacement target: replace"]
    PriPath["Name itself: keep"]
    Blank["Neither: flag for manual review"]
    Name --> Check
    Hint -.-> Check
    Check -->|secondary candidate| SecPath
    Check -->|primary candidate| PriPath
    Check -->|no match| Blank

Disambiguation with context (label / id / xref)

Alias hints are one kind of context: a per-row piece of independent evidence that helps decide which entity an ambiguous name actually means. update_ids and update_labels support three kinds, via ContextSpec:

• label -- an alias/synonym string (the synonyms=/--synonyms shown above).
• id -- a related/foreign identifier string, matched the same way.
• xref -- a cross-reference token (e.g. an Ensembl ID) resolved through an independent crosswalk table (XrefMapping).

All three only ever touch cells already flagged ambiguous, and every attempt can be written to an auditable decision log:

from pysec2pri import generate_labels, load_xref_mapping, update_labels

label_ms = generate_labels("hgnc")
ensembl_to_hgnc = load_xref_mapping("ensembl_to_hgnc.tsv")  # subject_id/object_id/object_label

resolved = update_labels(
    df, label_ms, at="gene_name",
    xref="ensembl",                # column with each row's Ensembl ID
    xref_mapping=ensembl_to_hgnc,
    report_path="decisions.tsv",   # stage, token, predicate_id, candidate, accepted, reason
)

The same options are available on the CLI:

pysec2pri update-labels genes.tsv hgnc --at gene_name \
  --xref ensembl --xref-source hgnc_custom --xref-on ensembl \
  --report decisions.tsv

Crosswalk tables

--xref-source names a table listed in the source's config. hgnc_custom is HGNC download, one row per gene:

Two columns of it are already a crosswalk: pick Ensembl ID and HGNC ID and you can map one to the other with --xref ensembl --xref-source hgnc_custom --xref-on ensembl

Bringing your own table

Pass any table with --xref-file. It needs three columns: subject_id (what you key on), object_id (this source's identifier), and object_label (its label):

subject_id         object_id   object_label
ENSG00000121410    HGNC:5      A1BG
ENSG00000175899    HGNC:7      A2M

pysec2pri update-ids genes.tsv hgnc --at gene_id --xref ensembl \
  --xref-file my_crosswalk.tsv

In Python you can point at the columns instead of renaming them, so a file like HGNC download works like:

from pysec2pri import generate_ids, load_xref_mapping, update_ids

xref = load_xref_mapping(
    "hgnc_custom.tsv",
    subject_col="Ensembl ID(supplied by Ensembl)",
    object_col="HGNC ID",
    object_label_col="Approved symbol",
)
update_ids(df, generate_ids("hgnc"), at="gene_id", xref="ensembl", xref_mapping=xref)

Diffing mapping sets

diff compares two SSSOM files (e.g. two releases of the same mapping set) and reports added/removed/changed rows:

pysec2pri diff old.sssom.tsv new.sssom.tsv --datasource hgnc -o diff.tsv

Documentation

Full documentation: https://pysec2pri.readthedocs.io/

Supported Databases

License

MIT License. See LICENSE for details.