MagHelix™ Zebrafish Screening Platform
CRISPR Model Pipeline 
Automated High-Content Screening 
AI Phenotypic Analytics | CADD/ADMET IntegrationMost preclinical teams face a binary choice: cell assays that miss organism-level liability, or mammalian studies that consume 6–12 months and $500K+ per program. The MagHelix™ Zebrafish Screening Platform eliminates that trade-off. We operate a fully integrated infrastructure — husbandry, genome editing, automated dosing, high-content imaging, and AI analytics — under one roof, with direct data handoff to our CADD and ADMET platforms.
Creative Biostructure at a Glance
Over a decade of trusted expertise powering biotech, pharma, and research institutions worldwide to advance therapeutic innovation.
Why Partner With Us
Most biotechs cannot justify the $2M+ CapEx of a zebrafish facility: recirculating aquaculture systems, IACUC infrastructure, transgenic breeding colonies, and high-content imagers. Pharma teams often fragment model generation, compound screening, and data analysis across separate CROs — losing the continuity required to link phenotype to mechanism. We built this platform to eliminate that friction: one infrastructure where CRISPR editing, automated phenotyping, and AI analytics share the same database and milestone clock.
Your CapEx is in chemistry and target biology. Ours is in zebrafish infrastructure and AI analytics.
| Stage | What We Deliver | What You Don't Need to Build |
|---|---|---|
| Model Generation | CRISPR/Cas9 knockout/transgenic lines; chemical induction models; breeding colonies | Vivarium, IACUC, microinjection suite |
| Compound Screening | 96-well automated dosing; multi-organ toxicity and efficacy readouts | Liquid handlers, imagers, behavioral trackers |
| AI Analytics | Deep-learning phenotypic scoring; behavioral pattern recognition; transcriptomic pathway mapping | Bioinformatics team, GPU cluster |
| Data Integration | Direct handoff to CADD, ADMET, and Lead Optimization | Computational chemistry infrastructure |
| Regulatory Package | GLP-aligned study reports; audit trails; dose-response documentation | QA/regulatory writing team |
Production-Ready Deliverables: Every campaign ships with phenotypic datasets, AI-quantified analytics, dose-response curves, and regulatory-formatted reports — ready for IND submission or investor diligence.
- ✓ Milestone-based pricing aligned with your fundraising cycles
- ✓ No vivarium overhead — husbandry, breeding, dosing, imaging, and analysis under a single project manager
Multi-organ toxicity. Transgenic disease modeling. Mechanism deconvolution. "Undruggable" is our starting point.
Proven platform scale
10,000+ compounds screened annually across cardiotoxicity, hepatotoxicity, neurotoxicity, and developmental toxicity endpoints — with mammalian concordance validated by peer-reviewed correlation studies (Spearman rho >0.85 for rat inhalation and oral acute toxicity).
Multi-modal integration
When a compound fails cell viability but shows selective rescue in zebrafish, our platform pivots to transcriptomic mechanism-of-action studies (10x Genomics Chromium, KEGG/GO enrichment) without restarting the project clock — data feeds directly into our AIDD model retraining.
IP firewall & encrypted data infrastructure
Full audit trails, client-isolated project folders, and contractual exclusivity on all custom transgenic lines.
Core Service Modules
Service Module At-a-Glance
| Service | Core Capability | Platform Integration | Typical Timeline |
|---|---|---|---|
| Zebrafish Disease Model Generation | CRISPR/Cas9 knockouts; transgenic reporter lines; chemical induction | AI-driven phenotypic baseline profiling; direct link to Structural Biology for target validation | 2–4 weeks |
| Zebrafish Toxicity & Safety Assays | Multi-organ liability detection; 96-well immersion dosing; automated imaging | ML anomaly detection; correlation with ADMET predictions; mammalian-predictive data output | 3–7 days |
| Zebrafish Efficacy Testing | Disease-relevant rescue assays; combination therapy matrices; high-content readouts | Bayesian dose-response modeling; direct handoff to Lead Optimization and CADD | 5–14 days |
Model Generation Pipeline
CRISPR Genome Editing at Platform Scale
Key Features:
- Automated Microinjection — Sutter CRISPR microinjection system with >90% targeting efficiency for standard loci; precision editing mimicking human disease polymorphisms.
- Transgenic Breeding Colonies — SPF-grade husbandry in Tecniplast ZEBTEC recirculating systems; automated temperature, light, and pH control ensuring batch consistency.
- AI Baseline Profiling — Deep-learning algorithms quantify subtle morphological and behavioral deviations across generations, replacing subjective visual inspection with statistical rigor before compound exposure begins.
What We Offer: For virtual biotechs, this eliminates the need to build a vivarium, hire husbandry FTEs, or procure microinjection infrastructure. For pharma, our platform generates bespoke models for targets lacking commercial assay kits — with genotyping reports and breeding protocols delivered under full contractual exclusivity.
High-Throughput Toxicity Screening
Multi-Organ Liability in a Single 96-Well Plate
Key Features:
- Comprehensive Endpoint Panel — Survival, morphology, cardiotoxicity (automated heart-rate imaging), hepatotoxicity (fluorescence transgenics), neurotoxicity (behavioral tracking), and developmental teratogenicity — all from a single embryo per well.
- ML Anomaly Detection — Noldus DanioVision behavioral tracking combined with deep-learning locomotor pattern analysis flags sub-lethal neurotoxicity that survival curves miss; automated image segmentation measures organ fluorescence and morphometrics with micron-level precision.
- Mammalian Predictivity — Platform-validated correlation with rodent acute toxicity (rat inhalation LC50: Spearman rho = 0.87; rat oral LD50: rho = 0.85) ensures zebrafish data serves as a reliable surrogate for mammalian risk assessment.
What We Offer: For lead optimization teams, this module provides an early warning system for off-target toxicity before expensive rodent GLP studies — conserving precious API with <10 mg compound per assay arm. For natural product programs, whole-organism safety profiling validates extracts at single-compound throughput.
Efficacy Validation & CADD Integration
From Phenotype Rescue to Mechanism-of-Action
Key Features:
- Disease-Relevant Assays — Angiogenesis (Tg(Fli-1:EGFP)), neuroprotection (MPTP/6-OHDA Parkinson's models), cardioprotection (doxorubicin challenge), and immunomodulation — quantified by high-content imaging with deep learning feature extraction.
- Combination Therapy Screening — Matrix dose-response designs evaluating synergistic or antagonistic effects in vivo, impossible to replicate in cell-based monoculture.
- Closed-Loop CADD Handoff — Efficacy data feeds directly into our CADD Platform for SAR model refinement; ADMET predictions inform dose-range selection before the first embryo is exposed.
What We Offer: When Hit-to-Lead biophysical data needs in vivo confirmation, our platform provides vertebrate proof-of-mechanism in weeks rather than months for rodent pilot studies. For difficult targets like PPI modulators, whole-organism pharmacokinetics reveal efficacy liabilities that cell-based docking scores cannot predict.
Platform Assay Portfolio
One Platform, Seven Therapeutic Areas, Twenty-Plus Validated Assays
The MagHelix™ Zebrafish Screening Platform hosts a comprehensive menu of disease-specific and organ-specific assays — all executable under standardized husbandry, dosing, and AI analytics protocols. The following matrix maps our validated assays to therapeutic applications, enabling rapid campaign design without assay development delays.
| Therapeutic Area | Assay / Model | Readout | Platform Integration |
|---|---|---|---|
| Angiogenesis | Normal angiogenesis assay (Tg(Fli-1:EGFP) larvae) | Intersegmental vessel formation; sprouting quantification | Parallel HUVEC proliferation, adhesion, migration, and tube formation assays for cross-species validation |
| Regenerative angiogenesis assay (Tg(Fli-1:EGFP) adult fin regeneration) | Fin ray vessel regrowth; re-epithelialization rate | Direct handoff to Wound Healing and Tissue Repair programs | |
| Toxicity & Safety | General toxicity on survival rates (embryo / larvae) | LC50; mortality curves; hatching rate | Correlated with ADMET predictions; feeds into In Vitro ADME-Tox Profiling |
| Cardiotoxicity test | Heart rate; QT interval; AV block; arrhythmia incidence | Automated video-cardiography; zERG channel homology mapping to human hERG liability | |
| Neurotoxicity (motor neuron / dopaminergic neuron development) | Locomotor activity; TH+ neuron count; axonal length | ML behavioral tracking; whole-mount IHC quantification | |
| Neurological Disorders | Protective effect against H&sub2;O&sub2;-induced PC12 / SH-SY5Y cell damage | Cell viability; oxidative stress markers | Cross-validated with zebrafish in vivo neuroprotection for mechanism triangulation |
| Protective effect against L-glutaric acid (LGA) induced neuron cell death | Neuronal survival; apoptosis (TUNEL) | Chemical induction model for excitotoxicity screening | |
| Protective effect against MPTP / 6-OHDA induced dopaminergic neuron death (Parkinson's disease) | TH+ neuron count in ventral diencephalon; DLS-validated α-synuclein aggregation | Transgenic and chemical Parkinson's models with validated positive controls | |
| Protective effect against MPTP / 6-OHDA induced locomotor deficit (Parkinson's disease) | Swimming velocity; distance; freezing time; light-dark preference | Automated Noldus DanioVision tracking with AI pattern classification | |
| Protective effect against chemical-induced epilepsy | Seizure-like behavior frequency; latency to onset; mortality rescue | Pentylenetetrazol (PTZ) and kainate models; EEG-like video analysis | |
| Cardiovascular Disease | Protective effect against chemotherapy drug-induced cardiac damage | Heart rate recovery; ejection fraction proxy; pericardial edema incidence | Doxorubicin and trastuzumab challenge models; ex vivo heart culture compatibility |
| Protective effect against chemical-induced cerebral hemorrhage | Hemorrhage incidence; hematoma size; survival | Collagenase and chemical hemorrhage stroke models | |
| Skeletal Disease | Bone anabolic effect | Bone mineralization (Alizarin red / Alcian blue); vertebral column length; AI skeletal anomaly scoring | BEiT deep learning classification (68.1% accuracy, 84.3% AUC) for subtle skeletal phenotypes |
| Immunology | Immunosuppressive effect on T-cell | T-cell proliferation; cytokine expression; thymus morphology | Fluorescence reporter lines; transcriptomic immune pathway profiling |
| Gene Expression | mRNA expression marker (per gene) | qPCR / RNA-seq quantification of target and pathway genes | Single-cell RNA-seq (10x Genomics Chromium) for cell-type-specific expression mapping |
Why this matters for platform users: Every assay in this matrix has been validated against mammalian reference data where applicable, and all share the same husbandry, dosing, imaging, and AI analytics infrastructure. For virtual biotechs, this means accessing a pre-built assay menu without assay development overhead. For pharma, this means mixing and matching endpoints across therapeutic areas under a single project manager — from a cardiotoxicity screen on Monday to a Parkinson's locomotor rescue assay on Thursday, with data delivered in a unified format.
Cardiovascular Toxicity — A Platform Highlight
Cardiovascular toxicity remains a leading cause of drug candidate attrition. A key platform advantage is the zebrafish heart's capacity for rapid dissection and ex vivo maintenance for several days — enabling acute and chronic cardiotoxicity assessment in a single vertebrate system. Combined with automated heart-rate imaging and zERG channel homology mapping, this assay delivers mammalian-predictive cardiac liability data at a fraction of rodent cost and timeline.
Technology Platform
Integrated Zebrafish Infrastructure: Husbandry + Genome Editing + Automated Screening + AI Analytics, Zero Handoffs
Traditional in vivo screening splits model creation, compound dosing, and phenotype analysis across separate vendors. Our platform unifies all stages under one roof, with AI predictions informing dosing strategies and every screening result feeding back into model optimization and computational model retraining.
Computational Platform — Dry Lab
| Capability | Details |
|---|---|
| AI Phenotypic Scoring | Deep learning image analysis (BEiT, ResNet, ViT) for morphological anomaly detection; automated behavioral classification with >95% inter-rater reliability |
| Behavioral Analytics Engine | Noldus DanioVision-compatible tracking; swimming velocity, distance, freezing, and light-dark preference quantification |
| Dose-Response Modeling | Non-linear regression and benchmark dose (BMD) analysis for efficacy and toxicity thresholds |
| Transcriptomic Integration | 10x Genomics Chromium single-cell and bulk RNA-seq; KEGG/GO pathway enrichment; GSEA for mechanism deconvolution |
| CADD/ADMET Handoff | Direct data pipeline to Molecular Docking, FEP, and ADMET Prediction platforms |
Experimental Platform — Wet Lab
| Capability | Details |
|---|---|
| Husbandry & Breeding | Tecniplast ZEBTEC multi-rack recirculating systems; SPF-grade embryo production; automated water quality monitoring |
| Genome Editing | Sutter CRISPR microinjection system; Tol2 transposase-mediated transgenics; morpholino knockdown; >90% standard locus targeting efficiency |
| Automated Dosing | Tecan Fluent liquid handler; 96-well immersion dosing; microinjection for poorly soluble compounds; 384-well compatible for combination matrices |
| High-Content Imaging | Zeiss Axio Observer Z1 fluorescence microscopy; PerkinElmer Operetta CLS for multi-channel time-lapse organ-specific phenotypes across 96-well plates |
| Behavioral Tracking | Noldus DanioVision high-throughput arenas; automated locomotor and oculomotor phenotyping across 96 larvae simultaneously |
| Molecular Analysis | 10x Genomics Chromium for single-cell/bulk RNA-seq; whole-mount immunohistochemistry; TUNEL apoptosis staining |
Tecniplast ZEBTEC
Tecan Fluent
Zeiss Axio Observer Z1
Noldus DanioVision
Platform Edge: The ability to generate a transgenic line on Monday, dose 96 compounds on Tuesday, and deliver AI-quantified phenotypic rescue data by Friday — all under one project team — compresses traditional 3-month in vivo validation into 2-week iterations.
Platform specifications are subject to continuous upgrade. Contact our team for instrument availability and project-specific capability assessment.
Closed-Loop Discovery Engine
When Phenotype Meets Prediction
Static cell assays predict molecular binding. Our platform reveals the organism-level reality — and feeds every result back into the design cycle.
ADMET Risk Prediction
AI toxicity models and virtual screening prioritize compounds and flag cardiotoxicity risk before in vivo exposure
→ Feeds into Dose Design
In Vivo Phenotypic Screening
Automated imaging and behavioral tracking quantify toxicity and efficacy across organ systems; transcriptomics link phenotype to pathway
→ Feeds into Mechanism Mapping
AI Analytics & CADD Refinement
Phenotypic datasets retrain docking scoring functions and ADMET models; SAR hypotheses updated with vertebrate PK/PD reality
→ Feeds into Lead Optimization
Structural & Computational Feedback
Experimental MOA data validates AlphaFold3-predicted binding sites and improves next-campaign target assessment
→ Feeds back into AI
Industrial Value:
For Biotechs
Your first zebrafish campaign's phenotypic data trains the AI models for your second compound series. Rescue data from Phase 0 becomes training data for Phase 1 — a compounding learning partnership.
For Pharma
Every phenotypic prediction is linked to an experimental outcome with project ID, timestamp, and model version — fully audit-ready for regulatory submissions and internal portfolio reviews.
Project Management & Execution
Project Workflow
A standardized, milestone-driven execution system. From model selection to regulatory-ready data package.
01 Strategy & Model Design
- Target review; model selection (CRISPR vs. transgenic vs. chemical); compound format confirmation; dose-range prediction via ADMET modeling
Deliverable: Study protocol + Gantt milestones + risk assessment
02 Model Generation / QC
- CRISPR guide design and microinjection; genotyping and phenotypic validation of F1/F2 generations; wild-type line QC confirmation
Deliverable: Validated model with genotyping report + QC documentation
03 In Vivo Exposure & Phenotyping
- 96-well compound administration; multi-parameter readouts (survival, morphology, organ fluorescence, behavioral phenotypes); real-time cardiotoxicity and neurotoxicity monitoring
Deliverable: Raw phenotypic dataset + interim QC report
04 AI Analysis & Transcriptomics
- Automated morphometric quantification via deep learning; behavioral pattern recognition; RNA-seq and KEGG/GO pathway enrichment for mechanism deconvolution
Deliverable: AI-processed phenotypic report + transcriptomic analysis + pathway maps
05 Regulatory Reporting & Transition
- Cross-endpoint statistical integration; dose-response modeling; regulatory-formatted study report; handoff to Lead Optimization or ADMET-Tox
Deliverable: Final technical report + raw data package + regulatory documentation
Sample Requirements
| Sample Type | Specification |
|---|---|
| Test Compounds | Powder or DMSO stock; minimum 10 mg per assay arm; MW and solubility data recommended for dose-range prediction |
| Compound Characteristics | Water-soluble or formulatable preferred; custom formulation (DMSO, ethanol, methylcellulose) available upon consultation |
| Data Input | Preliminary in vitro ADMET or biophysical data welcome for targeted endpoint selection |
| Model Preference | Transgenic vs. chemical induction; developmental stage (embryo vs. larva vs. adult) |
Standard Deliverables
Upon project completion, clients receive comprehensive experimental reports including:
- Validated disease model or toxicity/efficacy dataset with full QC traceability
- AI-quantified phenotypic analysis (morphometrics, behavioral metrics, organ-specific fluorescence)
- Transcriptomic profiling report with KEGG/GO pathway enrichment and mechanism hypotheses
- Statistical analysis with dose-response curves and confidence intervals
- Regulatory-formatted study report suitable for IND-enabling documentation
- Full electronic data package (raw images, sequencing files, analysis scripts)
- Direct handoff to Lead Optimization, ADMET Prediction, or In Vitro ADME-Tox Profiling
Our technical team responds within 24 hours. All inquiries protected under NDA.
Frequently Asked Questions
Case Study
Case: AI-Assisted Phenotyping in a Zebrafish Hypophosphatasia Model Enables Early and Precise Detection of Skeletal Alterations
Goal: Demonstrate the industrial application of AI-integrated zebrafish phenotypic screening for rare disease drug discovery.
Published Evidence:
Hark R, et al. AI-assisted phenotyping in a zebrafish hypophosphatasia model enables early and precise detection of skeletal alterations. Sci Rep. 2025.
Key Findings (from literature):
- AI Model Performance: BEiT deep learning model achieved 68.1% accuracy and 84.3% AUC in classifying zebrafish skeletal phenotypes into genotype classes (wildtype, heterozygous, homozygous) — significantly outperforming human manual scoring (38.0% accuracy, near random guessing).
- Explainable AI: Attention rollout visualization revealed the AI detected clinically relevant features (otolith morphology) that human observers overlooked, demonstrating unbiased pattern recognition.
- Drug Screening Application: The AI model enabled rapid, scalable evaluation of treatment efficacy by classifying whether compound-treated homozygous larvae phenotypically resembled wildtype — providing a quantitative rescue endpoint for high-throughput screening.
Industrial Translation:
We deploy BEiT-based phenotypic classification alongside our automated imaging pipeline on the MagHelix™ platform. For biotechs targeting rare skeletal disorders or other diseases with subtle phenotypic variation, this means access to AI-augmented vertebrate validation without the cognitive bias and throughput limitations of manual scoring. For pharma, our AI-quantified phenotypes provide audit-ready, quantitative datasets that support both regulatory submissions and internal portfolio prioritization — compressing the traditional phenotype assessment bottleneck from weeks to days.
Figure 1. Correlation of AI analyzed structures by attention rollout visualizations overlaid on the original microscopic images. (A.1 & B.1) Annotated illustrations on the left side indicated key structures of the (A) neurocranium and (B) viscerocranium.
Need platform-scale zebrafish screening to validate your preclinical pipeline? Our team can design a customized zebrafish platform deployment tailored to your target class, compound properties, and regulatory milestones. Contact our scientific team today.
