LoginSign Up
← Back to all sparks
M

MIC

INFRA · APIS
Velocity0.0

Analysis of Antimicrobial Minimum Inhibitory Concentration Data

github.com/agerada/mic ↗

An antimicrobial testing package cut its genomic half loose to become one thing well.

antimicrobial-resistanceclinical-microbiologycensored-datamethod-validationpackage-split
◆Current state
MIC analyses and validates antimicrobial minimum inhibitory concentration data. Version 2.0.0 removed every genomic function and ported them to a separate package, faLearn, leaving MIC focused on concentration analysis and method validation. The result needs fewer dependencies and no longer requires compilation, which is a meaningful install-time change for a clinical microbiology audience.
◆Where it's heading
The split was announced a release before it happened, with the genomic functions deprecated in 1.2.0 and the destination named. Around that, the substantive work has been on censoring: MIC values are reported with inequality signs, and 1.1.0 then 1.2.0 built out increasingly precise handling of them, separating less-than-or-equal from strictly-less-than and adding tolerance arguments to the comparison and agreement functions. That is the package's real subject matter, and shedding the genomic code is what makes room for it.
◆Prediction
With the scope settled, expect continued work on validation semantics and on the AMR package interoperability the notes already track. The relationship between MIC and faLearn is the thing to watch, since users of the removed functions now have to install both.

◆Recent moves

  1. 8mo ago

    Genomic functions removed and ported to faLearn

    ⚡ SPARK

    The release that finishes the split announced one version earlier. MIC becomes a single-purpose package for concentration analysis and validation, and the machine-learning-on-genomes work leaves for its own repository.

    View source ↗
  2. 9mo ago

    Genomic functions deprecated; censoring distinguishes leq and geq

    The release that pre-announced the split, deprecating the genomic functions and naming faLearn as their destination. It also sharpens the censoring model that is the package's core concern, distinguishing less-than-or-equal and greater-than-or-equal from their strict forms, with leq and geq arguments on force_mic and tolerate_ variants on the comparison functions.

    View source ↗
  3. 1y ago

    essential_agreement rebuilt around explicit censoring control

    The first of the censoring releases, replacing implicit handling in essential_agreement() and compare_mic() with tolerate_censoring and tolerate_matched_censoring arguments and defaults aimed at the common case of comparing an investigational method against a gold standard. Supporting methods for subsetting and dropping unused levels arrive alongside, plus compatibility with AMR v3.0.

    View source ↗