LoginSign Up
← Back to all sparks
K

KLINK

ANALYTICS
Velocity2.5

Kinship Analysis with Linked Markers

github.com/magnusdv/klink ↗

Forensic kinship testing with linked STR markers, now peer-reviewed and taking data from anywhere.

forensic-geneticskinship-testingstr-markersshinyr-packages
◆Current state
KLINK is a Shiny application and R package for kinship testing that accounts for linkage between STR markers, rather than treating all markers as independent. The built-in genetic map covers 50 common STR markers, pairs them dynamically after data loads, and lets the user set the maximum distance at which a pair counts as linked. Custom marker maps can now arrive as Excel files including KLINK's own download files, and a paper describing the tool was published in FSI:Genetics in 2026.
◆Where it's heading
The direction is from fixed panel toward general instrument. Hard-coded linkage pairs gave way to dynamic pairing, the built-in map grew from 18 to 50 markers, the map itself was then moved out to the norSTR package, and the app now accepts custom maps in several file formats and decimal conventions. Method capability tracks the upstream pedsuite rather than KLINK itself: special lumping in pedprobr and pedmut is what allowed complex mutation models in common pedigrees, and the newest release pins pedtools 2.11.0 and pedprobr 1.1.0 for faster likelihoods and better handling of looped pedigrees. Much of the remaining work is unglamorous casework ergonomics, replacing stray Norwegian text in Excel output, controlling table heights, fixing karyogram colours.
◆Prediction
With the map externalised and custom maps loading from multiple formats, the likeliest next step is broader marker-panel coverage through norSTR rather than changes in KLINK itself. Method gains should continue to arrive as pedprobr and pedmut version bumps.

◆Recent moves

  1. 12d ago

    Excel marker maps, more robust map loading, and a published paper

    Custom marker maps can now be Excel files, including KLINK's own download files where the map sits in the first sheet, and comma-decimal files load correctly, which matters for the European labs this tool is aimed at. The pedtools and pedprobr bumps bring faster likelihoods and better looped-pedigree handling, following the pattern where KLINK's method gains arrive from upstream. Publication in FSI:Genetics gives casework users something citable, which for forensic software is part of the product.

    View source ↗
  2. 3mo ago

    In-app help pages, a mutation-model setting, and clearer controls

    Interface work aimed at unsupervised use: help buttons and doc pages, a setting for mutation models, relabelled and relocated controls, and an improved karyogram. The Excel output cleanup, removing leftover Norwegian words and an obsolete report sheet, is a small sign of the tool generalising past its original lab.

    View source ↗
  3. 1y ago

    Complex mutation models become usable via upstream special lumping

    Adopting the special lumping methods from pedprobr and pedmut lets stepwise and other complex mutation models run in paternity and sibling pedigrees, with nonlumpable models falling back to the simpler Equal model in larger ones. The capability is upstream work rather than KLINK's, but this is the release where it reaches casework. Moving the default map out to norSTR starts the externalisation the later releases build on.

    View source ↗
  4. 1y ago

    Download fix for XML initials containing hyphens

    A single-line bug fix in XML input handling. No behaviour change beyond the failing case.

    View source ↗
  5. 1y ago

    Unlinked report picks the higher-PIC marker from each pair

    Marker selection for the unlinked report now favours the higher PIC value within each pair, computed consistently against one database. A defensible tie-break rather than a change in what the tool can do.

    View source ↗
  6. 2y ago

    Markers pair dynamically against a 50-marker map, replacing hard-coded pairs

    ⚡ SPARK

    The release that turns KLINK from a fixed-panel tool into one that adapts to whatever data a lab loads. Everything after it, the externalised map, custom map formats, the distance setting, extends the parameterisation introduced here.

    View source ↗