Hotdish: software for integrated analysis and scaffold-based hit triage of high-throughput screens

Scaffold-keyed triage for high-throughput screening: Four-stage Hotdish workflow with a screening funnel: plate-gradient correction of raw heatmaps, hit and interference selection on a replicate scatter plot, clustering of scaffolds into chemotypes as a similarity graph, and a chemotype-to-scaffold-to-compound triage view — narrowing 50,000 compounds screened in two repeats to 507 reproducible hits on the RyR2 biosensor, 144 chemotypes pooled across both RyR1 and RyR2, and one chemistry-informed retest list.

Chemotype-level triage across two screens: The Hotdish triage interface: a four-level drill-down from chemotype to generic scaffold to Murcko core to individual compounds, showing hit counts side by side for two ryanodine-receptor screens. Paired bar charts reveal a subset of ten compounds that are inactive on RyR2 and strongly active on RyR1.

IP Status: Unregistered Copyright; Pending U.S. Trademark Application (Serial No. 99527554)

Applications

• Academic and industrial high-throughput screening (HTS) data analysis
• Medicinal chemistry and chemical biology research
• Educational instruction in assay development and cheminformatics courses
• Laboratory-wide hosted or single-user local screening analytics

Key Benefits & Differentiators

• Unified Workflow Architecture: Integrates data loading, artifact correction, spectral fitting, multidimensional visualization, and chemical clustering within a single interactive interface.
• User-Friendly Accessibility: Eliminates programming or data science prerequisites for researchers analyzing complex microplate assay datasets.
• Scaffold-Based Chemotype Clustering: Reduces each compound to its Bemis-Murcko scaffold and its generic form and clusters those generic scaffolds into interactive chemotypes using RDKit and the Butina algorithm. Grouping on a shared chemical core rather than by pairwise compound similarity keeps related compounds together and maintains chemical scaffold membership across screens.
• Cross-Campaign Scaffold Namespace: Scaffolds carry a shared identity across every screen processed. Hits from one campaign can be read against all other campaigns, surfacing promiscuous cores and untested analogs of interest, all from primary data.
• Automated Correction and Interference Filtering: Parallelized per-well data fitting algorithms feed robust (median/MAD) Z-scores, optional per-plate gradient correction, and a Similarity-Index filter to flag interfering compounds.
• Flexible Deployment: Deploys as a Docker-hosted browser instance serving an entire laboratory, with per-user project isolation, or runs locally from a Python environment on a single researcher’s machine. Hosted users need nothing installed beyond a modern browser.
• Reproducible Analysis: Every parameter and decision, including gradient correction method, hit thresholds, clustering cutoffs and filters, etc., is written to an append-only per-project log so that any result can be traced back to the settings that produced it.

Technology Overview

High-throughput screening (HTS) and microplate-based assay workflows in drug discovery frequently suffer from fragmented data processing pipelines requiring multiple disparate software tools. Researchers commonly depend on spreadsheet software, which fails to scale effectively for high-volume HTS datasets, proves highly inefficient, and introduces significant human error that impacts data reproducibility. While commercial assay management tools exist, they are often cost-prohibitive and proprietary. Conversely, existing open-source analytical tools generally lack user-friendly interfaces, requiring advanced command-line scripting or data science expertise that creates substantial bottlenecks during compound triage and structure-activity relationship (SAR) analysis.

Researchers at the University of Minnesota have developed a Python-based software application named Hotdish™ that unifies HTS assay data organization, multidimensional visualization, and chemical clustering into a single interactive environment. Built upon core open-source scientific computing libraries including pandas, NumPy, RDKit, and HoloViews, Hotdish introduces domain-specific workflow automation tailored for primary drug discovery screening. It reads raw fluorescence-lifetime and spectral plate-reader data and fits per-well decay and spectral metrics. It also accepts pre-fit, per-well values from any other assay type in tabular format. The platform features parallelized data fitting, per-plate gradient correction, robust statistics for hit selection, and a Similarity-Index interference filter. Its central design choice is to organize the primary hit list by chemical scaffold rather than by compound, so structure-activity relationships become visible and inform retest decisions during the first round of triage without needing to rely on time-consuming and expensive follow-up assays. By delivering these advanced capabilities through an intuitive graphical interface, Hotdish enables researchers to triage primary hits by chemical class without writing custom code.

Phase of Development

TRL: 4-5
Fully functional application with a complete graphical interface, versioned releases, and a published user guide. Deployed both as an access-controlled hosted instance and as a local install, and validated on a published 50,000-compound fluorescence-lifetime FRET campaign screened against two ryanodine-receptor isoforms.

Desired Partnerships

This technology is now available for:

• License
• Sponsored research

Please contact our office to share your business’ needs and learn more.

Press Releases

University of Minnesota Publication Date

Researchers

• David Thomas, PhD Professor, Department of Biochemistry, Molecular Biology, and Biophysics
• Andrew Thompson Research Associate, Department of Biochemistry, Molecular Biology, and Biophysics