Logo
Homepage
Explore Our Models
My Cart
Contact
Subscribe
Models
Our Products
MouseAtlas
iPSC Cell Lines
Knockout Cell Lines
Tumor Cell Lines
Adeno-associated Virus (AAV) Standard Capsid
Featured Catalog
Humanized Mouse Models
HUGO-Ab™
Target-Humanized Mouse Models (HUGO-GT™)
Target-Humanized Mouse Models (CDS)
Humanized Immune System Mouse Models
Tool Mice
Cre Mouse Lines
Disease Models
Autoimmune Disease Models
Ophthalmic Disease Models
Immunodeficient Mouse Models
Metabolic Disease Models
Neurological Disease Models
Oncology & Immuno-oncology Models
Custom Model Services
Model Generation Techniques
Turboknockoutᵀᴹ Gene Targeting
Cre-ESCs Gene Editing
Targeted Gene Editing
Genetically Engineered Animals
Knockin Mice
Knockin Rats
Knockout Mice
Knockout Rats
Transgenic Mice
Transgenic Rats
Transgenic Model Generation
Virus Packaging
Adeno-associated Virus (AAV) Packaging
Adenovirus Packaging
Lentivirus Packaging
Custom Cell Line Services
Induced Pluripotent Stem Cells (iPSCs)
Knockout Cell Lines
Knockin Cell Lines
Overexpression Cell Lines
Point Mutation Cell Lines
Breeding & Supporting Services
BAC Modification
Breeding Services
Cryopreservation & Recovery
Phenotyping Services
Preclinical Solutions
Antibody Discovery Platform
HUGO-Mab™
HUGO-Light™
HUGO-Nano™
HUGO-Ab-eKO™
Therapeutic Area
Neurology
Alzheimer's Disease (AD)
Parkinson's Disease (PD)
Huntington's Disease (HD)
Blood Brain Barrier (BBB)
Neuropathic Pain
Metabolic
Obesity
MASH & MASLD
Cardiovascular
Ophthalmology
Glaucoma
Age-Related Macular Degeneration (AMD)
Retinal-related diseases
Oncology
PBMC Humanized Mouse Model
Human Immune System (HIS) Mouse Model
Immunology & Inflammation
Asthma
Drug Classification
Therapeutic Antibody Drugs
Monoclonal Antibodies (mAb)
Bispecific Antibodies (BsAb)
ADC/AOC
AI-Powered AAV Discovery
Cell Immunotherapy
Gene Therapy
Oligonucleotide Therapy
Fully Human Antibody Library
Neurology Antibodies
Metabolic & Cardiovascular Antibodies
Ophthalmology Antibodies
Oncology Antibodies
Immunology & Inflammation Antibodies
Resources
News
Blogs & Insight
Promotion
Events & Webinars
Databases
AbSeek
Rare Disease Data Center
Cell iGeneEditor™ System
iCyagen
Citations
Resource Vault
OriCell
About Us
Animal Health & Welfare
Company Overview
Facility Overview
Our Team
Our Partners
Careers
Health Reports
Contact Us
Login
HomeNewsCyagen Researchers Introduces IgScope, a Multi-Resolution AI Model for Antibody Humanness Assessment
November 12, 2026

Cyagen Researchers Introduces IgScope, a Multi-Resolution AI Model for Antibody Humanness Assessment

The region-isolated antibody language model evaluates humanness at the chain, region, and residue levels, offering a new computational approach to more detailed antibody sequence assessment.

Guangzhou, China, October 30, 2026 - Cyagen Biosciences has introduced IgScope, an AI-powered antibody language model that enables multi-resolution assessment of antibody humanness at the full-chain, regional, and individual amino acid levels. By independently evaluating antibody variable-domain regions, IgScope addresses an important challenge in antibody humanness assessment: distinguishing regional sequence characteristics that may be obscured by overall scores. While existing tools offer different levels of sequence analysis, IgScope integrates chain-, region-, and residue-level assessment within a unified framework, with a region-isolated architecture designed to evaluate local sequence features independently.

The model provides a computational approach to assessing AI-generated antibody sequences, supporting antibody humanization workflows, and evaluating sequences derived from human antibody and engineered chVHH discovery platforms.

A Region-Isolated Approach to Humanness Assessment

IgScope is a 23-million-parameter antibody Transformer model trained on 119 million human, mouse, and camelid antibody sequences. Its architecture combines two complementary modules:

  • Region-Isolated Encoder: Restricts attention interactions to sequences within the same IMGT-defined region, enabling regional assessment of CDRs with reduced dependence on conserved framework-region information.
  • Global Encoder: Analyzes full antibody sequences and incorporates ancestral germline prediction (AGP) and mutation process prediction (MPP) as auxiliary tasks.

The model generates three levels of assessment in a single inference:

  • Chain level: Humanness scores for complete heavy-chain, light-chain, or VHH sequences.
  • Region level: Individual scores for the seven IMGT-defined regions (FR1–CDR1–FR2–CDR2–FR3–CDR3–FR4).
  • Residue level: Humanness probabilities for individual amino acid positions.

Model Evaluation Across Multiple Datasets

The study evaluated IgScope through independent dataset testing, residue-level analysis, and comparison with clinical anti-drug antibody (ADA) data.

  • Regional independence: When framework-region tokens were masked, CDR-level identification accuracy remained largely unchanged. In a separate back-mutation benchmark, IgScope achieved 93.4% accuracy in identifying mouse-derived residues introduced into human antibody frameworks.
  • External dataset evaluation: IgScope achieved an AUC-ROC of 0.9997 for species classification on an independent VDJServer dataset and an ROC-AUC of 0.9938 for distinguishing species-origin patterns in 69,899 in silico-generated chimeric antibody sequences.
  • Clinical ADA association: Across a dataset of 217 therapeutic antibodies with reported clinical ADA data, IgScope's chain-level humanness scores showed a negative correlation with ADA incidence (Pearson r = –0.567). In a retrospective classification analysis using a 10% ADA-incidence cutoff, the model achieved a ROC-AUC of 0.807.

These results demonstrate the model's ability to characterize antibody sequence humanness across multiple evaluation settings. However, sequence-based predictions should not be interpreted as standalone or prospective clinical predictions of immunogenicity.

Applications in Antibody Discovery and Quality Assessment

1. Evaluating AI-Generated Antibodies

The study analyzed 3,359 antibody sequences generated by four AI models: BoltzGen, IgGM, ODesign, and RFantibody. IgScope's regional analysis provides an additional approach to identifying CDR sequence features with lower humanness scores that may be obscured by overall chain-level assessments.

This finding highlights the limitations of relying solely on overall scores to assess AI-generated antibodies. IgScope's regional analysis provides an additional approach to identifying potential non-human sequence features in CDRs.

2. Assessing Fully Human Antibody Mouse Platforms

Cyagen's HUGO-Ab™ platform includes HUGO-Mab™ fully human monoclonal antibody mice, HUGO-Light™ fully human common light chain antibody mice, and HUGO-Nano™ human-framework VHH mice with camelid hallmark features.

To help maintain VHH stability, HUGO-Nano™ chVHH sequences intentionally retain four camelid-characteristic amino acids (FERF) in the FR2 region. IgScope's showed a localized reduction in the FR2 humanness score, while the remaining regions retained high scores. In a computational masking analysis, masking the four retained residues restored the FR2 score to a human-like level.

This illustrates the potential of multi-resolution scoring as an in silico sequence-assessment approach for distinguishing intentionally retained engineering features from other sequence characteristics associated with lower humanness scores.

Limitations and Future Development

The current model evaluates individual antibody chains and does not account for heavy-light chain pairing, VH–VL interface interactions, or three-dimensional structural information. Its ability to distinguish species origin may also be limited when corresponding human and mouse CDR motifs are identical. Furthermore, clinical ADA outcomes are influenced by multiple factors beyond antibody sequence, including dosing regimens and assay methods. IgScope's sequence-based predictions should therefore be considered complementary to, rather than a replacement for, experimental immunogenicity assessment.

Cyagen plans to integrate IgScope into its AbSeek™ intelligent antibody computing platform. Combined with sequencing data from the HUGO-Ab™ platform, the integration is expected to support AI-assisted evaluation of antibody sequences, including AI-generated and humanized candidates, and contribute sequence-based humanness assessment as one component of early-stage developability evaluation.

Research Reference

Title: Multi-resolution humanness scoring of antibodies via a region-isolated language model

DOI: https://doi.org/10.64898/2026.10.23.761395

About Cyagen Biosciences

Founded in 2006, Cyagen Biosciences is an AI-enabled biotechnology company advancing biomedical research through proprietary gene-editing technologies, disease models, computational platforms, and integrated preclinical solutions.

Cyagen's antibody discovery capabilities combine the HUGO-Ab™ fully human antibody mouse platform with the AbSeek™ intelligent antibody computing platform, integrating in vivo antibody generation with AI-driven sequence analysis, screening, optimization, and validation.

Through these integrated capabilities, Cyagen aims to help biopharmaceutical innovators prioritize promising candidates, streamline early-stage screening, and advance therapeutic antibody development.

Media Contact

Cyagen: marketing@cyagen.com

Subscribe to Receive Updates & Promotions From Cyagen
Subscribe
* Your privacy matters to us. We never share it with third parties.
Latest Updates
Cyagen Researchers Introduces IgScope, a Multi-Resolution AI Model for Antibody Humanness Assessment
November 12, 2026
View Details
Cyagen and Tsinghua Research Team Develop PPBind-3D/1D AI Framework for Protein–Protein Binding Affinity Prediction
November 02, 2026
View Details
Accelerating the Path to IND: Cyagen Empowers Global Biopharma with Integrated Preclinical CRO Solutions
September 08, 2026
View Details
Cyagen's ProF Humanized Mouse Platform Supports FDA IND Clearance for Chimigen's Cancer Immunotherapy
July 28, 2026
View Details
Share
Top
Ready to Elevate Your Research?
Discover how Cyagen can support your research. Let’s start a conversation.
Model Library
Model Library
Resources
Resources
Animal Quality
Animal Quality
Get Support
Get Support
Address:
2255 Martin Avenue, Suite E Santa Clara, CA 95050-2709, US
Tel:
800-921-8930 (8-6pm PST)
+1408-963-0306 (lnt’l)
Fax:
408-969-0336
Email:
inquiry@cyagen.com
Services
HUGO-GT™HUGO-Ab™iPSC Cell LinesAdeno-associated Virus (AAV) Standard Capsid
Drug R&D
NeurologyMetabolicOphthalmologyOncology
About Us
Animal Health & WelfareCompany OverviewOur TeamHealth Reports
Social Media
Disclaimer: Pricing and availability of our products and services vary by region. Listed prices are applicable to the specific countries. Please contact us for more information.
Copyright © 2026 Cyagen. All rights reserved.
Privacy Policy
Site Map
Stay Updated with the Latest from Cyagen
Get the latest news on our research models, CRO services, scientific resources, and special offers—tailored to your research needs and delivered straight to your inbox.
Full Name
Email
Organization
Country
Areas of Interest
Main Area of Research