Skip to content
HPLC QC LabMAST30034 · revival

03 · Focus Area 4 · Clustering HPLC measurements

Find the odd runs

Each sample injection produces a handful of numbers: retention time, peak areas, widths, plate counts. With no labels to learn from, the question is whether injections fall into natural groups, and which ones sit outside every group.

This page follows the team's FA4 pipeline: drop control injections and identifier columns, fill blanks with 0, standard-scale, project to two dimensions, cluster, and tune by silhouette. Projections run in a Web Worker so the page stays responsive.

Choose data and methods

Assay (non-control injections)

Projection

Clustering

2021 recommendation for AAE:

t-SNE map · DBSCAN · 300 injections × 6 features

Computing the t-SNE map for AAE…
Computing t-SNE in a Web Worker · 0%
review prompt

Hover, or focus the map and use ← → to step through injections left to right and ↑ ↓ to jump between review prompts.

Silhouette

–

sklearn definition, noise counted as a label

Clusters

–

K-Means

Review prompts

–

Planted found

–

Possible only because the data is synthetic

Review prompts

Injections the clustering puts outside or at the edge of every group. They are prompts for a person to check, not verdicts: a missing value filled with 0 can look just as odd as a degraded sample.
Review prompts for the current clustering
Injection NameSequence NameWhy it is here