EN KR
Linkedin Facebook

SaaS

Neoantigen Automatic Report System
(NEO-ARS®)

NEO-ARS® is a comprehensive service that identifies tumor neoantigens and covers the whole process from NGS data analysis to 3D structure-based prediction of peptide-MHC (pMHC) binding using multiple AI algorithms.
Top 20 immunogenic neoantigen candidates are ranked as promising targets for cancer immunotherapy.

NEO-ARS® is a comprehensive service that identifies tumor neoantigens and covers the whole process from NGS data analysis to 3D structure-based prediction of peptide-MHC (pMHC) binding using multiple AI algorithms.
Top 20 immunogenic neoantigen candidates are ranked as promising targets for cancer immunotherapy.
SCROLL DOWN

Solution

Cancer vaccine developers want to obtain 3D structural information about how genetic mutations occurring in self-peptides yield immunogenic peptides that stably bind to MHC molecules.

NEOscan™ predicts both pMHC binding energy and T cell reactivity based on known crystal structures

Many clinicians and biotech companies interested in personalized cancer immunotherapy hope to discover neoantigen candidates as promising targets for cancer vaccines and T cell therapies.

NEO-ARS® i) determines the patient's HLA type, ii) prioritizes genetic mutations considering several parameters (e.g., sequencing depth, VAF(%), RNA expression level), and iii) identifies neoantigen candidates. It covers everything from genome profiling to in silico analysis of the 3D binding structure of pMHC complex.

NEO-ARS® : Tumor-specific neoantigen prediction pipeline

A. Workflow

A. Workflow A. Workflow
A. NEO-ARS consists of three steps A. NEO-ARS consists of three steps

B. Selection of best binding pose

B. Selection of best binding pose B. Selection of best binding pose

C. External validation

C. External validation C. External validation
  • (A) NEO-ARS® consists of three steps.
    Step1: NGS data analysis at both DNA and RNA level for detection of somatic mutations. Tumor-specific antigen (TSA) score is calculated.
    Step2: The 3D docking structures of pMHC complex are investigated and the binding energies are calculated.
    Step3: T cell reactivity score is calculated by analyzing changes in intermolecular binding patterns.
  • (B) 5,000 conformers are generated for each peptide. The best binding pose of pMHC complex is selected using 3D-CNN and analyzed via MD simulation.
  • (C) NEOscan™ outperforms other tools. The ROC AUC evaluated using immunogenic peptides and non-immunogenic peptides in external datasets indicates reliable prediction model performance.

Application

  • Development of personalized cancer vaccine targeting patient-specific neoantigens
  • Development of off-the-shelf cancer vaccine or TCR-T cell therapy targeting shared neoantigens (custom service)
  • Analysis of neoantigen burden as a predictive biomarker for response to cancer treatment (custom service)
References:
  • Cho et al. “Development of the variant calling algorithm, ADIScan, and its use to estimate discordant sequences between monozygotic twins” Nucleic Acids Research. 2018 Sep 6;46(15):e92.
  • Systems and methods that provide information for personal neoantigen-targeted immunotherapy using big data of molecular dynamics based on AI model. Korea Patent No. 10-2406699. 2022.
  • Ka et al. “HLAscan: genotyping of the HLA region using next-generation sequencing data” BMC Bioinformatics. 2017 May 12;18(1):258.

About NEO-ARS®

NEO-ARS®, an AI-powered neoantigen discovery platform

NEO-ARS®, an AI-powered neoantigen discovery platform, utilizes 3D protein structure and physics principles to enhance accuracy in predicting tumor-specific T cell epitopes. Unlike existing algorithms relying on amino acid sequences, NEO-ARS® analyzes multiple binding poses through Molecular dynamics simulation in a 3D environment.

This approach enhances accuracy in selecting neoantigen candidates by assessing their potential binding affinity to MHC class I molecules and the probability of T cell priming. This high-accuracy neoantigen prediction technology is crucial for the success of cancer immunotherapies, such as vaccines and T cell therapies, addressing the challenge that only a small percentage of tumor-specific mutations are immunogenic.

3D structure of peptide-MHC complex

(TCR-KRAS G12D peptide-HLA-A*02:01)

NEO-ARS® predicts peptide-MHC binding pattern of immunogenic neoantigen based on 3D structure.

  • 1,000 binding pose simulations
  • Best pose prediction & freeenergy calculation
  • Prediction of physical propertiesof 3D binding complex by deep learning
  • Neoantigen Peptides
  • Conventional Tool
  • NEO-ARS®
  • Include tomor-specific mutation & Bind to MHC-I

    Conventional Tool

    NEO-ARS®

  • Presented on cell surface

    Conventional Tool

    NEO-ARS®

  • Accessible to T Cell Receptor

    Conventional Tool

    -

    NEO-ARS®

NEO-ARS® Solution

NEO-ARS® neoantigen prediction process is applied in two steps: the first one is NGS pipeline; the latter is NEO pipeline.
NGS pipeline, NEO-ARS® is an optimized algorithm to analyze tumor specific somatic variants and HLA typing based on bioinformatics of genomic data integrated between whole exome sequencing (WES) and transcriptome sequencing.
NEO pipeline, AI-based prediction methods with 3D-CNN model for Peptide-MHC binding, key-residue model for TCR-peptide binding and MD simulation can predict physicochemical properties and dynamic T cell responses.
Our maximized neoantigen prediction platform provides a comprehensive solution between tumor mutations based therapeutic approaches.

In-vitro validation

NEO-ARS® predicts peptide and MHC binding based on 3D-CNN AI model and performs fine-tuning through MD simulations to provide more accurate and rigorous calculations. The selected candidate peptides are effectively narrowed down to those more likely to induce an immune response, selecting more meaningful candidates for subsequent experimental validation.

In-vitro validation
  • NEO-ARS® is a good predictor of MHC-peptide binding.
  • Neoantigen candidates were predictedusing NEO-ARS® v0.8.
    • Shared tumor mutations from publiccancer genome data
    • For common HLA types
    • Peptides with REVEAL score of ≥ 45% of the positive control are considered as good binders.

AI-model performance evaluationusing in vitro peptide-HLA binding assay (previous study 2022’)

  • NEO-ARS® is a good predictor of neoantigen-specific T cell reactivity.
  • Detection of T cell responses to patients’ private neoantigen candidates identified by NEO-ARS®
    • Tested by IFN-γ ELISpot assays with autologous peripheral blood samples
    • Samples from patients with kidney cancer or AML
    • Despite T cell exhaustion, neoantigen-specific T cell responses were observed

AI model performance

NEO-ARS® predicts peptide and MHC binding based on 3D-CNN AI model and performs fine-tuning through MD simulations to provide more accurate and rigorous calculations. The selected candidate peptides are effectively narrowed down to those more likely to induce an immune response, selecting more meaningful candidates for subsequent experimental validation.

Work Process

Work Process Work Process
Work Process Work Process
NGS data analysis process
  • Genomic profiling data sets of patients of interest (i.e., WES of tumor tissue and normal tissues, typically peripheral blood; RNA-seq of tumor tissue) provided by customer in fastq format.
  • NGS-ARS® is used for analyzing genomic profiling data. WES and RNA-seq data is analyzed individually to assess;
    (1) genetic variants of tumor tissue, (2) genetic variants of normal tissue, (3) HLA alleles, and (4) gene expression level. (5) High confidence, missense somatic (i.e., tumor-only) variants are then determined by comparing (1) and (2), cross-referencing with RNA-seq reads in tumor tissue, and quality assurance of sequence reads.
NEOscan™ prediction process
  • Input peptide sequences are generated. For each missense somatic variants, nine variant-containing g9-mer peptides and wild-type (WT) counterparts are generated. Note: Nine variant-containing peptides are generated per variant to test all possible positions of a given variant within the peptide.
  • NEOscan™ algorithm are run, and neoantigen candidates are determined;
    (1) In first module, peptide-MHC (pMHC) interaction is simulated in silico, which involved a series of computations of physicochemical properties of interaction at atomic-level. As result, best binding poses of pMHC – of 9 variant peptides and a WT peptide are determined. (2) In second module, the proprietary TCR interaction features are compared between the best pose of variant pMHC and that of WT pMHC to predict a likelihood of TCR binding. (3) Final list of neoantigen candidates are generated. Note: Predictions are performed separately for individual HLA alleles of the patient (3-6 HLA alleles per patient).

Terminology related to NEO-ARS® method

  • Term
  • Definition
  • Neoantigen candidate

    Definition

    The top 20tumor-specific antigens (TSAs) per patient, which are the most promisingtargets identified using NEO-ARS® for personalized cancer immunotherapy
  • B.E.

    Definition

    • Mean binding energy estimated using 10-ns of molecular dynamics (MD) simulation
    • Binding Energy (B.E.) is calculated by ENVA, an in-house package that calculates empirical bond energy based on Coulomb's law, which has the similar property as free energy in a high-throughput method.
  • T cell reactivity score

    Definition

    • Score indicating immunogenic potential based on calculation of structural differences in pMHC complexes between self and non-self
    • This is a normalized scoring system from 0 to 100, with peptides scoring >60 having a high potential to induce T cell responses

NEO-ARS®

This program is in-silico prediction tool, so we recommend proceeding with next-stage experiments such as immunogenicity or binding affinity tests with these neoantigen candidates

It covers the whole process from NGS analysis to neoantigen prediction.. NEO-ARS® neoantigen prediction process is applied in two steps: the first oneis NGS pipeline; the latter is NEO pipeline

Process

  • Input fastq or bam
  • WES analysis with paired data (tumor and normal tissue sequencing) and transcriptome analysis with tumor RNA sequencing
  • Extract tumor specific variants through DNA/RNA variant concordance step and to check the expression level of variants
  • Generate 9-mer peptide

Input

WES pairedtumor/
normal tissue,
RNA-seq with tumor
tissue,
Fastq or BAM

Runtime

~5 days per 1 run

Output

9-mer’s 20 neo-
peptide list

Support

Computer capacity :
I Unit of 1 CPU/1 GPU

send us your inquiry

SUBMIT
SUBMIT