Validating your assay - ScreenAssist
The Dianthus uHTS ScreenAssist App presents both overview and detailed plots on an interactive dashboard to allow a scientist to rapidly assess assay quality. Each plot tracks a different dimension of assay quality, including over time. The following section explains what each plot shows, what a good result looks like, and how to diagnose common problems.
Preparing your data for ScreenAssist
ScreenAssist allows rapid assessment of assay quality. As the Dianthus uHTS is ‘blind’ to enable full assessment, it is important to provide mapping information for your assay plates. ScreenAssist provides templates, but you are able to upload any plate mapping, either in .xlxs or in .csv format.
Currently you can upload one assay plate, either a single measurement, or a time course if you are monitoring assay stability over time.
Reviewing your data in the Overview Window
When a single assay plate is measured at multiple time windows to assess assay stability (i.e. a time course assay stability experiment), use the Overview Dashboard to assess at what point the assay provides the best signal to noise.
Plot 1: Spectral Shift Ratio Z-Prime Over Time
What it shows: Both Z-prime (Z’) and Robust Z-prime plotted against time for the plate. A horizontal pass threshold line is drawn at Z’ = 0.5.
What to look for:
The Robust Z’ should sit at or above the 0.5 pass threshold for the majority of plates.
Robust Z’ should be at or above classical Z’ at all times. If Robust Z’ is substantially higher than classical Z’ on a plate, this indicates outlier wells are dragging down the classical metric — the assay itself is likely fine, but outlier investigation is warranted.
A progressive downward trend in both metrics over time suggests signal drift, for example, protein degradation, dye instability, or a shift in assay conditions mid-screen.
An progressive upward trend over time suggest an assay stabilizing over time.
Acceptance thresholds:
Z’ / Robust Z’ |
Status |
≥ 0.5 |
🟢 Excellent |
≥ 0.4, < 0.5 |
🟡 Marginal |
< 0.4 |
🔴 Fail |
Common patterns and causes:
Pattern |
Likely cause |
Robust Z’ >> classical Z’ |
Outlier wells in the control population (autofluorescence, precipitation) |
Both metrics declining over time |
Protein or dye instability over the screen duration |
Both metrics consistently < 0.5 |
Insufficient assay window; revisit target/ligand concentration or dye labeling, perform a buffer optimization experiment. |
Plot 2: Spectral Shift Ratio SSMD Over Time
What it shows: The absolute Strictly Standardized Mean Difference (|SSMD|) between the positive control and neutral reference populations for each plate, plotted over time. A horizontal pass threshold line is drawn at |SSMD| = 6.
What to look for:
|SSMD| should remain at or above 6.
A gradual declining trend (as seen over a long time period) is expected and acceptable as long as values remain above 6. A sharp drop may indicate a sudden change in assay conditions.
SSMD complements Z’: a plate can have a borderline Z’ but an excellent SSMD if the two control populations have very different variances. Reviewing both together gives a more complete picture of assay quality.
Acceptance thresholds:
SSMD |
Status |
≥ 6 |
🟢 Pass |
< 6 |
🔴 Fail |
Common patterns and causes:
Pattern |
Likely cause |
SSMD declining across the screen |
Progressive assay signal drift -> check protein stability and dye lot consistency |
SSMD passes but Z’ fails |
Large variance asymmetry between positive control and reference -> investigate outliers in the higher-variance population |
Plot 3: Spectral Shift Ratio Distribution
What it shows: A probability histogram of the Spectral Shift ratio values for all wells across the screen, split into three populations: Reference, Control/positive control, and Ligand/compound wells. The x-axis is the Spectral Shift ratio value; the y-axis is probability density.
What to look for:
The reference and control populations should form **two distinct, narrow, well-separated peaks*. Good separation (with minimal overlap) is the visual equivalent of a high Z’ and SSMD.
The compound (ligand) distribution should be broad, spanning the space between, and potentially beyond, the reference and control peaks. This is expected: most compounds will be non-binders (clustering with the reference) and a small proportion will be binders (shifting toward the control peak).
The width of the reference and control peaks reflects assay variability (narrow = low CV and RSDE = good).
Any shoulder, secondary peak, or heavy tail on the reference or control distributions suggests outlier populations — autofluorescent compounds, aggregators, or quenchers that warrant further investigation.
Common patterns and causes:
Pattern |
Likely cause |
Broad or flat reference peak |
High Neutral RSDE; inconsistent dye loading or sample variability |
Overlapping reference and control peaks |
Low assay window; low Z’ and SSMD expected |
Heavy tail on the compound distribution |
Autofluorescent or quenching compounds contributing false signals |
Secondary peak within compound distribution |
A distinct population of strong binders, or a systematic artefact (e.g., DMSO effect at a specific concentration) |
Plot 4: Fluorescence % CV Over Time
What it shows: The percentage coefficient of variation (% CV) of the 650 nm and 670 nm fluorescence intensity signals for both the Reference and Control populations, plotted per plate over time. A threshold line is shown at 5% CV.
What to look for:
All four traces (Reference 650 nm, Reference 670 nm, Control 650 nm, Control 670 nm) should remain well below 5% throughout the time course.
CV values should be stable over time. A rising trend suggests increasing inconsistency in target stability over time.
The 650 nm CV is the primary acceptance criterion (pass threshold ≤ 5%). The 670 nm channel provides additional context on target photophysical stability.
Acceptance thresholds:
650 nm CV (Positive Control & Reference) |
Status |
≤ 5% |
🟢 Pass |
> 5% |
🔴 Fail |
Common patterns and causes:
Pattern |
Likely cause |
CV rising across the screen |
Progressive change in target stability over time. |
Control CV >> Reference CV |
Positive control compound affecting target stability (e.g., quenching, inner filter effect) |
670 nm CV elevated but 650 nm CV normal |
Dye photostability issue or 670 nm channel-specific optical artefact |
Plot 5: Fluorescence Change Over Time
What it shows: The mean absolute fluorescence intensity (A.U.) at 650 nm and 670 nm for the Reference and Control populations per plate, plotted over time. Error bars indicate the standard deviation (σ) within each population per plate.
What to look for:
The 650 nm and 670 nm signals for both populations should remain stable over time.
The Reference and Control populations should track together. If the two populations begin diverging in absolute fluorescence (not in ratio) over time, this suggests one population is being selectively affected (e.g., the positive control compound is affecting target stability).
Common patterns and causes:
Pattern |
Likely cause |
Steady upward drift in both populations |
Normal -> slight increases in dye signal over long incubations |
Control fluorescence lower than reference at 650 nm |
Positive control ligand quenching or competing with dye |
Plot 6: Affinity Over Time
What it shows: A dual-axis chart displaying the results of the titration (dose-response) curve measured as part of assay validation, plotted over time. The left y-axis (log scale) shows the fitted Kd and EC50 values (in molar units) for the positive control interaction. The right y-axis shows the relative fluorescence change (%) across the titration, which corresponds to the response amplitude. Error bars on Kd and EC50 reflect the uncertainty of the curve fit on each plate.
What to look for:
Kd and EC50 values should remain stable across all plates throughout the screen. A consistent, stable Kd over time confirms that the assay is stable over measured assay time.
The relative fluorescence change (response amplitude) should remain be stable over time and be ideally below 20 %.
Large error bars on Kd/EC50 on individual plates indicate poor curve fit quality, possibly due to insufficient S/N on that plate.
Common patterns and causes:
Pattern |
Likely cause |
Kd/EC50 drifting upward (apparent affinity decreasing) |
Protein degradation or target instability over screen duration |
Large error bars on curve fit |
Low S/N; insufficient positive control concentration or dye signal |
Kd and EC50 diverging from each other |
Curve fit quality issue; possibly a non-ideal dose-response shape on some plates |
Plot 7: Spectral Shift Ratio Robust Z-Score Over Time
What it shows: A scatter plot of the per-well Robust Z-scores of the Spectral Shift ratio for every well of every plate plotted as well index, grouped by plate. Horizontal threshold lines are drawn at 0±3 RSDE (where 99.7 % of reference wells should lie). Each dot represents an individual well. Control, Reference and Ligand are represented by different markers.
What to look for:
The reference population should cluster tightly around a Z-score of 0. This is expected, since the neutral reference is the normalization baseline. A large number of reference wells with Z-scores far outside ±3 are outliers (and should be investigated if spatially colocated).
The control population (orange dots) should cluster consistently at a high positive or negative Z-score, well away from the ±3RSDE threshold. This confirms that the positive control is reliably identified as a “hit” on every plate.
A shift in the control population’s Z-score cluster over time (e.g., drifting from +20 to +10) suggests the assay window is shrinking, consistent with declining S/N or response amplitude observed in Figure 6.
Common patterns and causes:
Pattern |
Likely cause |
Reference wells scattered widely around 0 |
High Neutral RSDE on those plates; inconsistent reference population |
Control population Z-scores declining over time |
Assay window shrinking; revisit positive control stability |
Isolated reference outliers far from 0 |
Contaminated or failed reference wells -> review poor reads and fluorescence CV |
Control and reference populations overlapping |
Very low Z’ expected; assay window insufficient for reliable hit calling |
Control cluster consistent and high throughout |
Healthy assay -> positive control reliably exceeds the hit-calling threshold on every plate |
Interpreting All Plots Together
No single plot tells the complete story. The following combinations are particularly diagnostic to monitor assay stability over time:
Observation across plots |
Conclusion |
Z-Prime plot: Z’ < Robust Z’ by a large margin and Z-Score plot: reference outliers present |
Outlier wells in the reference population -> remove outliers and re-assess classical Z’ |
SSMD declining and Affinity plot: response amplitude declining |
Assay window is shrinking progressively -> likely protein or dye stability issue |
CV plot: CV rising and Fluorescence Change plot: absolute signal drifting |
Signal instability -> check dispenser performance and dye lot |
Distribution plot: broad or bimodal reference peak and Z-Score plot: scattered reference cluster |
High Neutral RSDE -> the reference population is too variable for reliable hit calling |
All plots stable and within acceptance criteria throughout |
Excellent screen quality -> proceed with confidence to screening |