1 1
A P P L I C A T I O N N O T E
Figure 3: Assay Signal Quality. A. Top panel fluorescence CV and lower panel Robust Z-Prime for each plate. B.
Example plot of reference and control signal values over plate. Solid line is median, and dashed lines are 3 robust
standard deviation equivalents from the control. C. Scatter plot of Well Scan Outlier scores against Well Index Value.
Well types colored as in legend. The step at well index ca. 6100 is from Day 1 to Day 2 of the screen.
SINGLE-POINT SCREENING AND HIT IDENTIFICATION
Assay Quality. The Dianthus uHTS has a machine learning based algorithm which analyses each
well scan on the fly and provides an outlier score based on fit coefficients of a robust fit of the well
scan. The higher the score, the higher the likelihood of a 'bad read'. Bad reads can be caused by
several factors, from dust on the plate, to aggregates in the well (either from poor protein quality or
caused by compound aggregation). Internally, NanoTemper, uses a cut off of 3.331 to discard wells
that are considered bad (Figure 3C, Supplementary Figure 1).
Assay quality is monitored based on two major factors, percent coefficient of variance (%CV) of raw
fluorescence which provides details on the target dispense and if any significant issues happened in
plate preparation. If the coefficient of variation of the reference is lower than 5 % this is considered
acceptable. Assay performance and consistency is monitored by calculating Robust Z-prime scores
[14] plate wise, monitoring the separation between DMSO reference and positive control. Each plate
contained 4 columns of DMSO reference and 4 columns of 10 µM U0126.
