Annot

Lightweight Python helpers to annotate untargeted metabolomics feature tables for both LC-MS and GC-MS. The package ships with curated reference libraries (HMDB-derived LC formulas, an internal LC library with RT, and a GC library with RT anchors) and matching routines that return best hits plus candidate lists.

Installation

pip install git+https://github.com/BM-Boris/annot.git

Requires Python 3.8+, pandas, numpy, and tqdm (installed automatically).

Quick start

from annot import run_lc, run_gc

# LC-MS against HMDB-style masses
results_lc = run_lc(
    data="features_lc.tsv",
    sep="\t",
    mz_col="mz",
    ion_mode="pos",        # or "neg"
    lib="hmdb",            # or "annot" for RT-aware internal library
    save="annotations_lc.csv",
)

# GC-MS with RT anchoring
results_gc = run_gc(
    data="features_gc.tsv",
    sep="\t",
    mz_col="mz",
    rt_col="rt",           # retention time in seconds
    save="annotations_gc.csv",
)

Input expectations

• LC-MS: tabular file (CSV/TSV) with at least an mz column; add rt when using the RT-aware lib="annot" mode.
• GC-MS: tabular file with mz and rt columns (RT in seconds).
• Columns are selected by name via mz_col / rt_col; everything else is preserved and returned with annotations.

LC-MS annotation

• lib="hmdb" matches feature m/z to monoisotopic masses and common adducts (pos: [M+H]+, [M+Na]+, [M+NH4]+, [M+K]+; neg: [M-H]-, [M+Cl]-). Restrict adducts with adducts=["[M+H]+", ...].
• lib="annot" uses the internal LC library (annot/data/lc_lib.csv) with observed RTs; matches on both m/z and RT.
• RT shift handling: shift="auto" infers a shift using Glutamine (pos) or Glutamate (neg); provide a float (seconds) to override.
• Tolerances: mz_diff is a relative tolerance (default 5e-6), time_diff is relative RT tolerance (default 0.05).
• Output columns include: original columns, annotation (best hit), adduct, ppm_error, formula, distance (matching score), and candidates (all hits within tolerances).

GC-MS annotation

• Uses the bundled GC library (annot/data/gcms_lib.csv) with m/z and RT anchors.
• shift="auto" estimates RT shift from 4,4'-DDE anchors (falls back to 0 if not found); provide a float to override.
• Peaks are filtered into groups within time_range (seconds) and require ngroup members (defaults: 2.0 sec, 3 peaks). Groups with confirmed mz0 anchors are preferred.
• Tolerances: mz_diff (relative, default 5e-6) and time_diff (relative RT, default 0.05).
• Output columns: original m/z/rt, annotation, notes (e.g., anchor labels), and candidates sorted by distance.

Libraries

The package includes:

• annot/data/hmdb.csv - HMDB-based LC masses/formulas.
• annot/data/lc_lib.csv - internal LC library with RT/adduct info.
• annot/data/gcms_lib.csv - GC library with RT anchors.

Use path in load_library_data or lib in run_lc to point to custom files if needed.

Tips

• Keep input RT units consistent (seconds throughout).
• Lower mz_diff/time_diff to tighten matches; increase when instrument error is higher.
• Save outputs via the save argument to capture the annotated table as CSV.