Company

About Gradient BioTech

A multi-area scientific research platform

Gradient BioTech brings study management, reproducible compute pipelines, interactive result exploration, and analysis into one shared application across research domains. Researchers can move from raw data to insight without switching between disconnected notebooks, viewers, and job trackers. Computational biology, cardiology, neurology, oncology, pathology, immunology, infectious disease, and microbiology each get purpose-built workflows, while sharing the same users, organizations, and platform services. Heavy computation runs as asynchronous, pollable jobs, so long-running pipelines never block exploration. And where the platform offers AI-assisted interpretation, it cites the underlying computed metrics rather than generating new results, keeping every conclusion traceable back to the analysis that produced it.

Our mission

We build tools that help teams move from raw data to reproducible insight. Every analysis run records parameters, pipeline versions, status, and artifacts so work can be reviewed, repeated, and shared. We believe reproducibility should be the default, not an afterthought bolted on at publication time. By keeping data, methods, and provenance together in one workspace, we aim to shorten the path from question to evidence and make it easy for collaborators to build on each other's work with confidence.

Research areas

Work is organized into eight research areas. Each area has its own data types, pipeline workflows, and analysis views while sharing users, organizations, and platform services.

Computational Biology

Transcriptomics from single cells to spatial maps

The Computational Biology area is a web workspace for transcriptomic analysis. Upload single-cell, bulk, or spatial datasets; run guided QC, clustering, integration, annotation, differential expression, pseudobulk, trajectory, enrichment, and biomarker pipelines; explore results interactively; and assemble publication-ready figures — with full provenance for every step.

Study workspaces with datasets, sample metadata, design tables, and contrasts
AnnData ingest with assay-role inspection for counts, logcounts, scaled data, protein, and spatial coordinates
Single-cell RNA-seq QC: genes/cells detected, mitochondrial fraction, doublet scores
Normalization, scaling, PCA, UMAP, and Leiden/Louvain clustering
Scalable sketch analysis with backed AnnData, Zarr/Parquet artifacts, projected labels, dedicated results, and large-dataset guardrails
Batch integration with corrected PCA representations, integrated clustering, before/after comparison views, batch-mixing metrics, and confounding warnings
Reference-based cell type annotation with compatibility checks, per-cluster score heatmaps, confidence gaps, warnings, and supporting marker genes
Wilcoxon differential expression with ranked gene tables and volcano views
Pseudobulk differential expression with saved-contrast prefill, per-cell-type/group volcano views, sample aggregate review, and TSV exports
Cell-cycle scoring and graph-based pseudotime with UMAP overlays, assumption warnings, pseudotime-associated genes, and TSV exports
Single-cell CNV and malignant-cell evidence from expression-derived chromosome-scale signal, with reference QC, event calling, subclones, state calls, review overlays, and oncology handoff artifacts
Reference-based deconvolution with NNLS-estimated cell fractions and fit diagnostics
GO and pathway enrichment from differential expression results with ORA, ranked GSEA, ssGSEA/GSVA score summaries, custom run-scoped gene sets, and set-operation review
Biomarker discovery with bounded multi-method feature comparison, inline Open Targets disease evidence, and cross-validated classifier metrics
Co-expression network module explorer with hub genes, eigengene-trait correlations, lightweight hub-edge graph, and TSV exports
Coverage-based DEG result review with pairwise contrast details, disease-association evidence, and TSV exports
Spatial transcriptomics domain clustering, SVG overlays, neighborhood enrichment, condition-domain comparison, niche candidates, and TSV exports for Visium-class `.h5ad` datasets
Interactive spot viewer with gene expression overlays
Explore workspace: QC charts, linked gene inspection, UMAP expression panels, spatial preview, and exportable summaries
Analyze workspace for chaining QC, clustering, integration, annotation, DE, pseudobulk, trajectory, biomarker, and spatial steps
Interpret workspace for annotations, enrichment, gene sets, and methods provenance
Disease context lookup mapping a gene signature against ~1.8M target-disease associations across ~24,000 diseases, with overlap enrichment (Fisher's exact, BH-FDR), integrated association scores, evidence-source badges (genetic, somatic, clinical, literature, pathway, expression), and expandable per-gene matched evidence.
Multi-panel figure builder with PDF export
Supplement-ready TSV exports for DE, enrichment terms, term genes, biomarker, and spatial result tables
Run history with pipeline versions, parameters, and stale-output warnings
Grounded analytical interpretation tied to computed genes, pathways, and metrics
Guided bulk RNA-seq workflow with filtering, normalization controls, VST, DE, sample QC, PCA/sample-distance diagnostics, and TSV exports
Async job polling with downloadable artifacts
Shared organization and user model across all studies

Cardiology

Multimodal cardiovascular and physiological signal research

The Cardiology area is a web workspace for multimodal cardiovascular and physiological signal research. Upload ECG, RR intervals, blood pressure, PPG, and respiration recordings; run validated preprocessing and metrics pipelines; link data to subjects; compare cohorts; and generate interpretations grounded in computed metrics.

Experiment containers for cardiovascular signal studies
ECG upload in WFDB, CSV, and EDF waveform formats
ECG preprocessing with R-peak detection, RR intervals, and signal quality index
Time-domain, frequency-domain, and nonlinear HRV metrics
Advanced HRV: SDANN, DFA, entropy, fragmentation, PRSA acceleration/deceleration capacity, and Lomb/AR spectra
Full ECG analysis pipeline chaining preprocess and HRV in one run
ECG delineation of P, Q, R, S, and T fiducial landmarks from detected beats
Per-beat and recording-level morphology: PR, QRS, QT/QTc, ST level, and T-wave metrics
RR-only CSV upload for Holter and wearable exports
Arrhythmia burden metrics: irregularity percentage, pause burden, and beat flags
Sliding-window HRV with per-window metrics, quality/missingness, trends, and feature matrices
Electrodermal activity (EDA): tonic/phasic decomposition and skin conductance responses
Accelerometer activity: ENMO intensity, motion burden, and per-window activity
Cross-modal motion quality flagging ECG/HRV beats and segments during movement
PPG preprocessing with pulse peak detection and inter-beat intervals
Pulse rate variability (PRV) computed from PPG preprocess runs
PPG pulse-wave morphology: amplitude, rise/decay time, width, area, and dicrotic notch
Pulse transit / arrival time (PAT, PTT) from paired ECG and PPG recordings
Blood pressure variability, dipping classification, and MAP statistics
Arterial pressure (ABP) waveform: beat-level SBP/DBP/MAP, pulse pressure, and upstroke dP/dt
Respiration rate, breath timing, and quality summaries
Respiratory sinus arrhythmia (RSA) cross-modal coupling from ECG and respiration
Baroreflex sensitivity by sequence method and cross-spectral transfer function (gain, coherence, phase, latency)
Subject import and cohort metadata linking across recordings
Cohort comparison with group means, standard deviations, and statistical tests
Windowed-feature cohort comparison on per-subject medians and temporal variability (CV)
Batch HRV jobs across multiple datasets in one request
Research-only signal-derived risk stratification from cohort outcomes metadata
Interactive waveform explorer for QC and visual context, with an optional P/Q/R/S/T landmark overlay
AI interpretation with descriptive and mechanistic modes citing recorded metrics
Methods report generation from experiment bundles

Immunology

Immune cell states, signaling networks, and repertoire dynamics

The Immunology area is a research workspace for immune-system datasets and translational immunology studies. Build on computational-biology single-cell and spatial workflows, add immune-specific annotation and state scoring, organize disease cohorts, and run repertoire, cytokine, cell-cell communication, and ligand-to-target hypothesis analyses with reproducible run records.

Study containers for autoimmune, infectious disease, vaccine, allergy, and immunotherapy research
Immune composition summaries with population counts, proportions, group summaries, and cohort context
Reference-atlas immune label transfer with probability scores, confidence categories, competing labels, marker evidence, compatibility checks, and fine-grained immune hierarchy workflows
Immune state scoring for exhaustion, activation, effector function, cytotoxicity, and regulatory suppression
Cytometry event analysis with arcsinh transform, self-organizing-map prototype clustering, metaclusters, minimum-spanning-tree edges, marker heatmaps, and sample abundance summaries
CITE-seq, VDJ, spatial, ATAC, FCS, cytokine, bulk RNA-seq, and single-cell dataset metadata
CITE-seq RNA/protein integration with ADT background correction, joint coordinates, RNA/protein discordance, and CD-marker phenotype panels
scATAC tabular peak analysis with cell QC, TF-IDF/LSI coordinates, marker peaks, gene activity, peak-to-gene links, group coverages, and motif summaries
Disease context and cohort metadata for timepoints, medication, vaccination, infection history, and response labels
Canonical home for immune repertoire analysis with diversity, clone expansion, CDR3 length/property/k-mer summaries, V/J pairing, sample distances, barcode clone maps, clone occupancy, clonal bias, VDJ ambiguity flags, and public/private clones shared with oncology workflows
Canonical home for immune cell-cell communication and cytokine network methods shared with oncology
Complex-aware ligand-receptor scoring with permutation p-values, adjusted p-values, specificity scores, pathway aggregation, and sender/receiver role summaries, plus a differential-communication table across conditions (per-condition scores, log2 fold change, and condition-label permutation FDR), visualized as an interactive cell-type communication network (per-pathway filter, role-grouped nodes, responder vs non-responder edge coloring), a sender→pathway→receiver flow (Sankey), a pathway-activity heatmap (sender→receiver pairs × signaling pathways), and per-cell-type communication fingerprints (role, hub score, distinct out/in partners, dominant sending/receiving pathways) with graph-centrality hub analysis — betweenness (broker), eigenvector (influence), and signaling entropy (specialist vs broadcaster)
Communication-motif detection over the directed signaling graph — reciprocal loops (feedback when both arms share a pathway, cross-regulatory when each direction uses a different pathway), feed-forward loops, and 3-cycles, each pathway-annotated and ranked by its limiting-edge strength — plus a drug-targeting overlay that annotates ligand/receptor genes with Open Targets tractability (druggable modality: small molecule, antibody, PROTAC) and safety liabilities (research evidence, not a treatment recommendation)
Ligand activity inference workflow with sender/receiver selection, receiver target-gene sets, ligand rankings, ligand-target matrices, and ligand-receptor-target paths
Graph-based spatial immune workflows with neighborhood enrichment, TLS-like candidate criteria, cytokine/state gradients, region summaries, and spatial edge payloads, with an interactive tissue scatter map (cells by type and neighbor density) and a cell-type neighborhood-enrichment heatmap
Run history for immune analysis parameters, artifacts, model versions, and reproducible outputs
Disease evidence lookup mapping immune signature gene sets against ~1.8M target-disease associations across ~24,000 diseases, with overlap enrichment (Fisher's exact, BH-FDR), integrated association scores, datasource counts, and evidence-type provenance (genetic, clinical, literature)
Report generation with methods text, tables, figure specs, provenance, and grounded immune-analysis outputs
Grounded analytical interpretation for immune cell states, signaling, repertoire, and disease context

Infectious Disease

Infection episodes, epidemiology, spread intelligence, and cross-area evidence

The Infectious Disease area is a web workspace for episode-centered infectious-disease cohort research and population-level spread intelligence. Model participants, infection episodes, specimens, exposures, interventions, and outcomes; link versioned Microbiology, Computational Biology, and Immunology results by subject and episode; run person-place-time epidemiology, cohort comparison, time-to-event, genomic cluster and transmission-network investigation, and host-pathogen-outcome integration; and use the shared Spread workspace for immutable connector snapshots, national and HHS-region surveillance curves, choropleths, Rt estimation, probabilistic forecasting, wastewater early warning, and WHO Global Health Observatory country indicators — with AI-assisted interpretation grounded in computed metrics and association-only guardrails throughout.

Episode-centered data model: participants, encounters, infection episodes, specimens, exposures, interventions, and outcomes
Versioned research case definitions with adjudication and revision history
Longitudinal infection timelines with explicit missingness and uncertain intervals
Person-place-time epidemiology: attack rate, incidence, hospitalization, and case-fatality with denominators and ascertainment
Epidemic curves and stratification by time, place, demographics, and setting
Cohort comparison with risk ratio, odds ratio, and chi-square/Fisher testing
Kaplan-Meier time-to-event analysis with log-rank testing and explicit censoring
Host-pathogen-outcome integration linking versioned cross-area results by episode, with provenance preserved
Exposure and contact network analysis surfacing candidate transmission-hypothesis links from contacts, shared locations with overlapping windows, and shared events — with a pathogen genomic-relatedness overlay, never proof of transmission
Outbreak cluster investigation combining case timing, location and contact evidence, and pathogen genomic relatedness into candidate clusters, with an interactive network view
Investigator adjudication that promotes, rejects, or resets candidate clusters with a recorded rationale and preserves those decisions across re-runs
Directional transmission-hypothesis links with serial-interval and genomic-distance likelihood, link-level accept/reject/uncertain adjudication, and threshold sensitivity sweeps
Hospital movement and facility investigation views for within-facility spread hypotheses
Pathogen lineage dynamics over time from linked Microbiology genomic results (frequency, replacement, growth-advantage summaries)
Landmark treatment/resistance-outcome comparison that defuses immortal-time bias, stratifies by antimicrobial-resistance evidence linked from Microbiology, and flags confounding by indication — descriptive and non-prescriptive, never treatment effectiveness
Observational vaccine effectiveness (VE = 1 − risk/odds ratio) with confidence intervals, breakthrough-infection summaries, an optional test-negative design, and subgroup views — association only, never causal efficacy or a policy claim
Instantaneous reproduction number (Cori renewal method) with a caller-supplied serial interval and credible intervals, plus epidemic-growth rate and doubling time — a retrospective research estimate, never real-time surveillance
Spread intelligence workspace: shared connector registry, immutable content-hashed raw snapshots (GCS archival), evidence records with license/attribution, and Cloud Tasks connector refresh
Machine-enforced redistribution gate blocking raw export and bundling for restricted or unknown-license sources
CDC Open Data connectors: national NREVSS respiratory positivity and lab test/detection volumes; HHS Region 1–10 positivity trends; state RSV and SARS-CoV-2 centered 3-week positivity/tests and trailing 5-week detections with explicit eligibility/suppression gaps; NSSP state emergency-department visit percentages for RSV, SARS-CoV-2, influenza, and combined respiratory illness; NNDSS weekly state reports for measles, pertussis, hepatitis A, Lyme disease, West Nile virus, and salmonellosis with distinct no-cases, unavailable, and not-reportable states; state COVID/influenza hospital admissions and hospitalized census; U.S. Census population denominators
HHS-region Epidemic Dynamics analysis with stable regional geography, region-specific curves, Rt estimation, probabilistic forecasts, rolling evaluation, and per-region data-quality review
Surveillance data-quality panel: source freshness, geographic coverage, missingness, gap detection, and revision deltas between snapshot vintages
Comparison workspace for up to six regions, states, or countries on a shared measure scale, plus unit-safe multi-signal trajectory comparison using independently indexed laboratory, ED, hospitalization, and wastewater series with per-series missingness and snapshot provenance
Population-normalized epidemic curves with explicit date basis; series preflight guard refusing mixed date bases, structural breaks, or too-few-points before Rt/forecast runs
U.S. state choropleth map (deck.gl) with Census per-capita normalization and small-count suppression
Spread Rt estimation: raw vs delay-adjusted stable Rt with provisional tail flagging, Rt=1 change-point crossings, serial-interval × window sensitivity grid, and archived estimates diffable by data cutoff
Probabilistic surveillance forecasting: no-change/seasonal/trend baselines, renewal/GLM/state-space models, ensemble comparison, rolling-origin backtest (WIS/MAE, coverage, calibration), and promotion gate refusing models that fail to beat baselines or are miscalibrated; append-only forecast archive
Wastewater early warning: CDC NWSS national viral-activity signals (SARS-CoV-2, influenza A, RSV), revision-window snapshot retention, dual-source lag sweep vs clinical/lab reference, and rolling leading-indicator evaluation
WHO Global Health Observatory (first slice): OData connector transport, ISO 3166 alpha-3 country crosswalk, country indicators (DTP3, measles MCV1, polio, HepB3, TB incidence, malaria deaths, HIV new infections), global world-map choropleth (percent, per 100k, per 1k), country epidemic curves with full-history default, and annual forecast/evaluation horizons
Observed / estimated / projected outputs kept visibly distinct in Spread UI, exports, and interpretation (retrospective, non-operational surveillance — not a command-and-control system)
Grounded AI interpretation with associational guardrails against transmission, causation, diagnosis, and outbreak claims
Async job polling with downloadable artifacts and reproducible provenance

Microbiology

Microbial communities, pathogens, and antimicrobial resistance

The Microbiology area is a web workspace for microbial community and pathogen research. Upload feature tables or processed profiles; run QC, taxonomic and community profiling, alpha/beta diversity with PERMANOVA, compositional differential abundance, and functional pathway analysis; compare cohorts; and generate AI-assisted interpretation grounded in computed metrics — with reference-database provenance for every step.

Study workspaces with sample manifests, datasets, saved cohorts, and run history
Feature-table, taxonomy, tree, and pathway-table ingestion with reference-database provenance
Microbial QC: per-sample depth, feature retention, prevalence, and low-depth flags
Stacked relative abundance bar charts by taxonomy rank (phylum through species) with per-sample views
Alpha rarefaction curves for sequencing-depth adequacy assessment
Alpha diversity (Shannon, Simpson, observed features, Chao1) with group significance testing (Kruskal-Wallis, pairwise Wilcoxon)
Beta diversity (Bray-Curtis or Jaccard), PCoA ordination, PERMANOVA, and betadisper dispersion test
ANCOM-BC differential abundance with sampling-fraction correction, 95% CI, and BH correction; CLR + Mann-Whitney also available
Core microbiome analysis with an interactive prevalence x abundance scatter and live threshold sliders
Longitudinal community analysis: within-subject trajectories, first-differences, and a repeated-measures linear mixed-effects trend test
Functional pathway abundance and cohort comparison with measured-vs-inferred distinction
Grounded AI interpretation citing computed metrics, with associational (non-causal) guardrails
Async job polling with downloadable artifacts and reference-database versioning per run

Neurology

Brain connectivity, EEG, and fMRI in one neuroscience workspace

The Neurology area is a web workspace for brain connectivity, functional imaging, and electrophysiology research. Upload connectivity matrices, EEG, or fMRI data; compute graph-theoretic network metrics; explore interactive brain visualizations; compare cohorts with Network-Based Statistics (NBS); and generate AI-assisted interpretations grounded in computed outputs.

Study containers for connectivity, EEG, and fMRI projects
Connectivity matrix upload in CSV, NPZ, NumPy, and MATLAB `.mat` formats
Brain Connectivity Toolbox graph metrics: clustering, modularity, and efficiency
Small-worldness, characteristic path length, and centrality measures
Network heatmap, graph layout, and 3D brain views when atlas coordinates exist
EEG preprocessing: bandpass, notch, re-reference, and bad-channel handling
Optional ICA for artifact reduction in EEG recordings
Interactive EEG artifact review: per-channel QC, ICA component browser with reject/keep, and cleaned-recording apply
EEG band power, spectral entropy, and PLV/coherence/wPLI connectivity
Advanced multi-estimator EEG connectivity (imaginary coherence, PLI, debiased wPLI, AEC) with an estimator-agreement comparison
EEG microstate analysis with coverage, duration, transitions, and GEV
Time-frequency spectrogram statistics and band-power dynamics
Simplified EEG source localization summaries
iEEG/ECoG workflow: shaft-aware bipolar referencing, high-gamma band power, and peri-event high-gamma response
fMRI BIDS archive validation with missing-field warnings
fMRI motion QC: framewise displacement, DVARS, outlier scrubbing, tSNR, and subject-exclusion flags
Functional connectivity with AAL and Schaefer atlas parcellation
Confound regression and bandpass filtering for FC pipelines
Dynamic functional connectivity with sliding-window states
Task fMRI first-level GLM: BIDS events, HRF design matrix, per-condition activation, contrasts, and timing QC
Batch graph metrics jobs across labeled subject datasets
Network-Based Statistics (NBS) for group-level edge comparisons
Graph-threshold sensitivity sweep with a robustness report flagging which metrics survive the density choice
Predictive connectome modeling (NBS-Predict): cross-validated classification and regression, permutation testing, and edge-stability maps
Group metric comparison tables with effect sizes
Cluster-based nonparametric EEG statistics: sensor × band permutation tests with cluster-mass FWE correction
Multiplex network metrics across connectivity layers: overlapping strength, participation coefficient, and layer similarity
Cohort analysis workspace with group assignment and batch jobs
Reusable protocol templates that chain preset pipeline steps in one click
Organization-level custom atlas and montage registry with validation on upload
Report export with methods text, annotations, and run manifest
Grounded analytical interpretation for graph metrics, NBS, prediction models, and microstate outputs
Async neurology job polling with versioned pipeline runs
4D Brain Exploration Workspace: synchronized orthogonal slice viewer (axial, coronal, sagittal) with shared crosshair cursor and MNI/RAS coordinate readout from the NIfTI affine
NIfTI volume loading with automatic role detection (anatomical, functional, statistical map, label) from header intent codes and dimensionality
Manual window/level controls for anatomy brightness with auto-windowing default
Statistical map overlay on anatomical slices with per-colormap rendering (hot, viridis, plasma, autumn, cool, RdBu_r), positive and negative threshold controls, opacity slider, and a colormap legend
Synchronized atlas region label: Schaefer 100/200/400 parcellation lookup at the cursor MNI coordinate, updating as crosshairs move
Template cortical surface renderer in the 3D brain view: toggleable pial/white/inflated hemisphere meshes with sulcal-depth shading, aligned to the same MNI space as atlas-centroid connectivity nodes

Oncology

Tumor microenvironment, immune response, mutations, and outcomes

The Oncology area is a research workspace for tumor cohorts and translational oncology studies. Organize samples, clinical endpoints, mutation and repertoire data; run tumor microenvironment, immune phenotype, communication, ligand activity hypothesis, mutation, and survival workflows; and interpret computed results with provenance.

Study containers for tumor cohorts, treatment arms, response groups, and timepoints
Sample and clinical metadata management with patient/sample identifiers
Tumor microenvironment composition analysis from expression-derived cell states
Oncology communication product layer with TME sender/receiver classes, IO pathway axes, clinical metadata summaries, spatial boundary priorities, permutation-tested ligand-receptor significance (BH-FDR), and a differential-communication table across conditions (per-condition scores, log2 fold change, and condition-label permutation FDR)
Cell-cell communication visualization and network analytics: interactive communication network (per-pathway filter, role-grouped nodes), pathway-flow Sankey, and pathway-activity heatmap with responder vs non-responder coloring, plus per-cell-type communication fingerprints (role, hub score, distinct out/in partners, dominant sending/receiving pathways), graph-centrality hub analysis — betweenness (broker), eigenvector (influence), and signaling entropy (specialist vs broadcaster) — and communication-motif detection (reciprocal feedback/cross-regulatory loops, feed-forward loops, and 3-cycles, pathway-annotated and ranked by limiting-edge strength)
Drug-targeting overlay that annotates communication ligand/receptor genes with Open Targets tractability — flagging druggable targets by modality (small molecule, antibody, PROTAC) plus safety liabilities — so suppressive edges and motifs can be read as therapeutically actionable (research evidence from Open Targets/ChEMBL, not a treatment recommendation or clinical decision support)
Tumor-immune phenotype classification (immune-inflamed / immune-excluded / immune-desert) from signaling axes and CD8 spatial penetration, with transparent inflamed/excluded/desert axis scores and supporting pathway evidence (CXCL9/CXCL10, IFNG vs TGFβ, MIF, CAF) — a research readout, not a diagnostic
Spatial TME communication summaries with tumor core, invasive margin, stroma, necrosis, immune-excluded, and TLS-adjacent region context, including a communication network faceted by tissue boundary
Ligand activity immuno-oncology hypothesis workflow for tumor/stroma/myeloid ligands, immune receiver target programs, and ligand-receptor-target chains
Tumor-board-style IO communication and ligand activity highlights with research caveats and clinical metadata linkage
Immuno-oncology profiling with deconvolution, immune dysfunction-and-exclusion response labels, exhaustion, bulk TME interpretation, metadata group comparisons, and repertoire metrics
Immune dysfunction-and-exclusion response modeling with dysfunction, exclusion, myeloid/stromal suppression, IFNG, MMR/MSI-expression proxy, CD274, CD8/CTL, and therapy-context outputs
Malignant cell detection workflow consuming CompBio single-cell CNV handoff artifacts, with malignant population summaries, CNV event evidence, mutation overlap, subclone TME features, and survival/response feature exports
Mutation landscape analysis with TMB, SBS-6/SBS-96 signature exposure, focal/arm CNA summaries, known driver CNA annotations, oncoprints, pathway enrichment, differential mutation testing, and clinical group enrichment
Survival analysis with Kaplan-Meier, log-rank tests, Cox regression, and longitudinal trajectories
Cross-validated predictive biomarker modeling (classification/regression) from expression features with selectable random-forest / linear / elastic-net models, permutation testing, and top-biomarker ranking
Integrated cohort comparison across communication, immune/TME, mutation, CNA, signature, ligand activity, and survival evidence
Disease evidence lookup mapping mutation, expression, or biomarker gene sets against ~1.8M target-disease associations across ~24,000 diseases, with overlap enrichment (Fisher's exact, BH-FDR), integrated association scores, datasource counts, and evidence-type provenance (genetic, somatic, clinical, literature)
Grounded analytical interpretation tied to oncology pipeline outputs and cited metrics
Run history for pipeline status, parameters, artifacts, and reproducible outputs

Pathology

Computational pathology and whole-slide imaging research

The Pathology area is a web workspace for computational pathology and whole-slide imaging research. Upload tissue slides or region images; generate tile manifests for browser viewing; run tissue detection, cell segmentation, and spatial quantification; inspect overlays in the slide viewer; and track reproducible pipeline runs.

Study containers for whole-slide imaging and region-image projects
Slide upload for TIFF, OME-TIFF, PNG, JPEG, and scanner-format files
Multi-level tile generation with thumbnail and pyramid manifests for web viewing
Tissue detection producing masks, region bounding boxes, and focus QC summaries
Cell and nucleus segmentation with overlay PNG and structured cell tables
Pluggable segmentation engines: watershed splitting of touching nuclei with split-count and boundary-quality QC
Spatial quantification: cell density heatmaps across tissue regions
Nearest-neighbor distance statistics between detected cells
Infiltration phenotype scoring and region-level metric summaries
CSV and JSON exports for downstream statistical analysis
Interactive slide viewer with pan/zoom tile rendering, a minimap navigator, and synced side-by-side slide comparison
Viewer layers for tissue mask, cell overlay, and density heatmaps
GeoJSON and QuPath-compatible annotation import
OME-TIFF export and QuPath-native GeoJSON round-trip (class + color) for QuPath/Fiji/OMERO interoperability
Tissue microarray dearraying: detect cores into a labeled grid, flag missing cores, and a TMA map viewer with per-core tissue fraction
Stain color deconvolution for H&E and IHC (H-DAB) with per-stain channels and DAB-positive area quantification
Interactive annotation authoring: draw polygon/rectangle/point regions, classify (tumor/stroma/necrosis/…), move, delete, and export GeoJSON
Trainable cell classifier: learn cell classes from annotation regions with cross-validated accuracy, per-class metrics, and feature importances
Cell morphometrics (area, perimeter, circularity, eccentricity, solidity, axes) with an interactive measurement explorer: sortable tables, histograms, scatter plots, and CSV export
Study-level pipeline run history with linked artifacts
Cohort comparison of pathology biomarkers across sample groups
Reproducible pipeline versioning recorded on every analysis run
Async pathology job dispatch with queued, running, and complete states

How work moves through the platform

Choose a research area — Each domain has dedicated dataset types, pipeline workflows, and analysis views under one shared organization and user model.
Create a study — Organize datasets, sample metadata, subjects, design variables, and analysis runs in a reproducible study container.
Run pipelines — Launch domain-specific compute as asynchronous, pollable jobs — QC, statistical and compositional analysis, clustering and differential testing, cohort and longitudinal comparisons, network and signal processing, or image and spatial quantification — each recording its parameters, versions, and artifacts.
Explore and interpret — Review dashboards, interactive visualizations, and run artifacts specific to each research area, alongside grounded summaries that cite the recorded outputs behind every conclusion.

Report Builder

The Report Builder is a cross-area document composer for turning analysis outputs into a narrative your team — and collaborators — can review together. Drag panels from any research area's studies into one report: plots, cohort comparisons, interpretation summaries, uploaded images, and authored text or markdown blocks. Arrange and annotate them in a single canvas instead of stitching together exports from separate tools.

Reports live in your organization's workspace, so everyone on the team can open the same document, see the latest version, and contribute without passing around slide decks or screenshot folders. That shared context makes it easier to align on results before a lab meeting, hand off work between analysts, or prepare material for a PI or external partner. When you need to share outside the platform, export the report to PDF. Each panel stays linked to the source run behind it, so what you share remains traceable to the analysis that produced it.

Feedback and support

Need help, found a bug, or have a product idea?

Press

For press inquiries, contact press@gradientbio.tech.

Contact

Questions about Gradient BioTech may be sent to contact@gradientbio.tech.