Increasing Response Rates in Immuno-Oncology with Artificial Intelligence

There are multiple factors contributing to the immune system recognizing cancer and properly launching adaptive immune reaction against it. Ardigen research combines the expertise in Artificial Intelligence, Bioinformatics, Cancer Biology, Immunology and Microbiome to deliver technologies that lead to significantly increased response rates to immunotherapies.


Microbiome Analysis Platform The whole metagenome information for effective therapy design

By using the whole metagenomic information available, Ardigen Microbiome Analysis Platform demonstrates an innovative approach to the discovery and design of microbiome-based therapeutics and diagnostics.
Our in silico methods, in particular Artificial Intelligence algorithms, minimize the laboratory effort to identify and evaluate of new probiotic strains. Novel Live Biotherapeutic Products, therapies based on bacterial metabolites and microbiome-related biomarkers can provide significant medical benefit to patients in immuno-oncology and beyond.

DataPatient’s microbiome- desired phenotype Patient’s microbiome - unwanted phenotype LBP CandidatesTherapeuticMetabolitesBiomeMarkersMicrobiome Analysis PlatformBiomeScoutfunctional analysis of the metagenome BiomeSeerstrain phenotype predictorValue

BiomeScout functional metagenome analysis

BiomeSeer is a high quality tool for analyzing the differences in microbiome between groups of patients, such as the differences between responders and non-responders in the context of a particular therapy.

Ardigen has designed a novel approach to detecting differences between the groups that is aware of the high complexity of metagenomic samples. BiomeScout™ operates on all the genomic features detected in the sample. This approach enables the detection of multiple changes introduced by strains of the same species. Detection of specific genomic features makes an important step towards the full functional analysis of metagenomic samples. It facilitates causative studies of clinically important metagenome functions. Ardigen approach to microbiome analysis provides, among others, the following benefits:

  • database bias free identification of differentiating genomic features that goes beyond the species level
  • reliable detection of patterns spread across different strains
  • gate to the functional analysis - a key to understanding how microbome interacts with the host
  • detection of novel bacteria strains with complete genomic information that facilitates its potential use as a Live Biotherapeutic Product
  • ability to predict and improve the patient’s response to the therapy
BiomeScoutValueDataPatient’s microbiome- desiredphenotypePatient’s microbiome - unwantedphenotypeMining for Genomic FeaturesDifferentiatingGenomic FeaturesGenomes of strainscarryingGenomic FeaturesAssembly ofGenomic Featurescarriers

BiomeSeer strain phenotype predictor

BiomeSeer is a tool for predicting bacterial phenotype, that allows to carry out the analysis in silico, eliminating most of the laboratory procedures. This element of the Microbiome Analysis Platform completes the functionality of BiomeScout, as it analyzes the assembled genomes of novel bacterial strains found in the samples.

BiomeSeer encodes genomic sequences into structures suitable for machine learning algorithms. Then, it performs a set of validation tests using custom functions designed to handle the biological bias. Finally, a number of pre-trained Machine Learning models are applied to the data. Together, they predict the most important aspects of the bacterial phenotypes, significantly reducing the effort necessary to establish the isolation and culturing protocols, including:

  • adhesion to the intestinal epithelium
  • survival in different conditions
  • antibiotics resistance
  • general biosafety
  • metabolites
DataNovel bacterialstrain genomePredicted phenotypeof the strainBiomeSeerMetabolitesOxygen requirementBiosafetyGram stainingAntibiotic resistanceValue

Neoepitope Prediction Platform Effective cancer vaccines design and TCR screening

Usually only a few of the tumor-related mutations create neoepitopes able to elicit an adaptive immune response. Predicting those neoepitopes is crucial to design effective therapeutic vaccine but this is an extremely challenging task.

Ardigen Neoepitope Prediction Platform can accurately predict cancer neoepitopes, assess their immunogenicity and design personalized cancer vaccines to boost the response to immunotherapy.

DataWhole Exome Sequencing- normal tissueWhole Exome Sequencing - tumorGene expressionsequencing - tumorImmunogenicneoepitopesPrioritized pHLAs for therapy developmentPersonalized vaccinecompositionNeoepitope Prediction PlatformArdImmune Rankneoepitope immunogenicityrankingArdImmuneVaxpersonalized cancer vaccine designValue

ArdImmune Rank neoepitope immunogenicity assessment

Ardigen's Neoepitope Prediction Platform applies Artificial Intelligence algorithms to the assessment of potential immunogenicity of peptides. It recognizes biological features that decide if a given peptide would be presented by the patient's HLA I complex and elicit adaptive immune response.

Recognizing the correct 1% of all the potential peptides that might be produced from mutations in a tumor and prioritizing them for laboratory work increases efficiency of pre-clinical R&D while reducing necessary budget and time. Ardigen's algorithms offer higher accuracy and PPV (Positive Predictive Value) than competing solutions.

ArdImmune Rank offers the functionality of ranking a list or database of short peptides by their potential immunogenicity, taking into account multiple factors, including:

  • specific HLA type
  • binding affinity and stability of HLA complex with given peptide
  • similarity to self-antigens
  • estimation of potential interaction with the T-cell receptor
  • repertoire
DataPeptide database - candidates for screeningArdImmune RankBinding affinity and stability to HLASimilarity to self-antigensImmunogenic neoepitopes for specified HLA typesPotential to attract T cell repertoireDelivered value

ArdImmune Vaxpersonalized cancer vaccine design

ArdImmune Vax is designed to support the demanding process of personalized cancer vaccine manufacturing. Tumors bear thousands of mutations but usually only a few of them create neoepitopes able to elicit an adaptive immune response. The tumor-related neoepitopes are presented by the HLA I complexes on the tumor cells and recognized by the T-cell receptor. This process can be boosted by the application of a vaccine training CD8+ and CD4+ cells to recognize tumor-specific pHLA complexes.

Cancer, however, is a very heterogeneous disease thus personalized approach for each tumor is poised to be more effective. Ardigen AI-based Neoepitope Prediction Platform uses the patient’s tumor and normal tissue sequencing information to predict the potential neoepitopes and assess their immunogenicity.

ArdImmune Vax can be used in the pre-clinical design and test phase as well in the highly time-constrained post-approval scenario. Our technology is aware of the complexity of neoepitope processing and factors contributing to it's immunogenicity. In particular, it takes into consideration the following:

  • not all somatic mutations of cancer give rise to neoepitopes
  • processes like peptide splicing generate novel epitopes
  • affinity and stability of the binding of peptide by the HLA drive peptide presentation
  • interaction of the HLA-peptide complex with T-cell repertoire is essential for adaptive immune response
  • similarity to self excludes a range tolerated or potentially dangerous epitopes
  • tumor clonal structure and frequency of neoepitopes influence effective clone elimination by cytotoxic T-cells
DataWhole ExomeSequencing- patients normaltissueWhole ExomeSequencing- patients tumorRNA Sequencing- patients tumorImunogenic neoepitopesfor patient's HLA types and tumor clonal structureComprehensive analysis reportVaccine neoepitopecombinationsArdImmune VaxNeoepitope predictionImmunogenicity scoringNeoepitope diversity scoreTumor clonal structure analysisValue

Biomarker Discovery Platform Maximizing the value of clinical trial data

Ardigen’s holistic approach to immuno-oncology analyzes the properties of tumor and its microenvironment, including composition of immune cells, biomarkers detectable in blood and other fluids, and microbiome composition. The data is used to build a comprehensive picture of a response to an immunotherapy and to select robust biomarkers to be used in the clinical trial assay and companion diagnostics development. Ardigen’s systems biology approach and cutting edge AI form a unique in silico platform for discovery of biomarkers predictive of immunotherapy response and facilitate the design of CTA and CDx.

DataRespondersNon-respondersOptimal assay definitionPatient stratification algorithm for specific therapeutic / indication pairBiomarker Discovery PlatformMulti-omics data integrationPublic and proprietary dataImmune signaturesSystem biology approachValueWGS, WES, Gene panelsRNA seqImmunohistochemistrySMSClinical parametersCustom data

Raw sequencing data preprocessing and custom biomarker integration

The Ardigen Biomarker Discovery Platform accepts various data types coming from multi-omics experiments, including raw sequencing of tumor and normal tissue, as well as an array of custom measurements. A range of fine-tuned preprocessing pipelines is applied to prepare high quality data for biomarker discovery modeling. The platform also accepts data that have already been processed and standardized.

  • Whole Exome Sequencing of tumor/blood pairs.
  • Variant Calling and Variant Annotation.
  • Gene expression quantification from RNAseq.
  • Normalization and standardization of proteomics, metabolomics, immunohistochemistry, cell phenotyping and other types of data.

Off-the-shelf immuno-signatures integration

Ardigen’s solution includes many ready to use genetic signatures described in peer-reviewed scientific publications to enable wider interpretation of patient stratification and benchmarking with public knowledge, including:

  • ESTIMATE (Yoshihara, Kosuke, et al. "Inferring tumour purity and stromal and immune cell admixture from expression data." Nature communications 4 (2013): 2612.)
  • TIDE (Jiang, Peng, et al. "Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response." Nature medicine (2018): 1.)
  • xCell(Aran, Dvir, Zicheng Hu, and Atul J. Butte. "xCell: digitally portraying the tissue cellular heterogeneity landscape." Genome biology 18.1 (2017): 220.)
  • IMPRES (Auslander, Noam, et al. "Robust prediction of response to immune checkpoint blockade therapy in metastatic melanoma." Nature medicine 24.10 (2018): 1545.)

Integrative immuno-oncology response modelling by AI

Ardigen’s Biomarker Discovery Platform implements AI algorithms that ‘understand’ different biological data and encode it in machine-readable fashion. That enables the integration and selection of measurements providing most valuable and comprehensive information that differentiates between responders and non-responders to an immunotherapy.

The unique prognostic and predictive abilities of our Biomarker Discovery Platform come from a set of rules that the system has been built around:

  • Focus on modeling response defined by clinical benefit.
  • Feature engineering driven with immuno-oncology knowledge.
  • Building models with low number of input variables.
  • Machine Learning best practices reflected in stratified sampling and intense cross-validation framework.
  • Predictive and prognostic evaluation of features and models using proprietary and public data sets.