Skip to content

About#

signatomics is a family of open-source Python tools for finding biomarker signatures in omics data, written by Elena Stamatelou. The first module, signatomics.olink, is for Olink (PEA) proteomics: Target 96/48, Explore, Explore HT and Reveal.

Who it is for#

Researchers who analyse Olink data in Python: biomarker and machine-learning studies, cohorts measured on several plates or projects, and anyone who wants each cleaning step to be visible and repeatable. If you work in R, Olink's own R package covers much of the cleaning.

Status#

Version 0.0.1, in development. Preprocessing (data import, quality control, bridging, sample metadata) and the first statistics (the analysis table, compare groups, correlate) work, with 104 automatic tests; prepare does the preprocessing in one call, and built-in synthetic data lets you try everything. It installs from GitHub until the first release on the Python package index; see Start. Function names may still change before the first release.

Roadmap#

  • Read NPX files: long and wide layouts; .csv, .txt, .xlsx and .parquet
  • Quality control: a quality check and charts, then your choices applied, with every removal reported
  • Bridging plates and projects, with independent checks
  • Sample metadata, with a report of every ID without a match
  • The usual route in one call (prepare), and built-in example data
  • The analysis table: one row per sample, one column per assay (to_wide)
  • Compare groups and correlate: one test chosen in advance, effect sizes with confidence intervals, FDR across all assays, volcano plots
  • An optional total-protein correction for samples whose dilution varies (saliva, urine)
  • Find a biomarker signature: nested cross-validation, stability selection, calibration
  • Check the signature: what each protein adds, learning and selection curves
  • Adjust for confounders; before/after and repeated measurements
  • First release on the Python package index, with a citable DOI
  • Later: ELISA and microbiome modules, and combining data types

Principles#

  • Nothing changes silently: every step that changes data says what it did.
  • Look before you act: checks and charts never change data.
  • Methods chosen in advance, before seeing the data.

Cite#

A DOI will come with the first release. Until then, cite the GitHub repository. If you use the public PAD000040 data, also cite the dataset and its paper (see the PAD000040 example).

Licence and trademarks#

An open-source licence (MIT) is planned for the first release. signatomics is an independent project, not affiliated with or endorsed by Olink or Thermo Fisher Scientific. Olink is a trademark of Thermo Fisher Scientific.