Computational Pipeline

Five-stage workflow from raw multi-omics data to patient-specific report
1
Multi-omics Data Acquisition
Collect and QC all input data types per patient

Inputs

  • WES/WGS (tumor + normal) — Illumina NovaSeq
  • Long-read RNA-seq (tumor) — PacBio Iso-Seq / ONT
  • Short-read RNA-seq (tumor + normal) — Illumina NovaSeq
  • Immunopeptidomics (tumor) — LC-MS/MS
  • scRNA-seq (tumor) — 10x Chromium
  • TCR-seq (blood/tumor) — ImmunoSEQ / 10x V(D)J
  • ATAC-seq (tumor, optional) — Illumina NovaSeq

Outputs

QC-passed FASTQ/BAM files Sample metadata sheet
↓
2
Bioinformatic Processing
Variant calling, alignment, HLA typing, AS/fusion detection, immunopeptidomics

Processing Steps

StepTool(s)
Somatic variant callingMutect2, Strelka2, FreeBayes
CNV callingCNVkit, FACETS
HLA typingOptiType, HLA-HD
HLA-LOH detectionLOHHLA, DASH
RNA-seq alignmentSTAR, HISAT2
Expression quantificationSalmon, RSEM
AS event detectionrMATS, SUPPA2, LongGF
Fusion detectionSTAR-Fusion, Arriba, LongGF
ImmunopeptidomicsMSFragger, FragPipe
scRNA-seq processingCell Ranger, Scanpy
TCR repertoireMiXCR, ImmunoSEQ Analyzer

Outputs

VCF files AS event catalog Fusion catalog Peptide identifications Cell type annotations Clonotype table
↓
3
Neoantigen Discovery & Prioritization
Translate variants, AS events, and fusions into ranked neoantigen candidates

Neoantigen Sources

SourceMethodBinding Prediction
Somatic mutationsVariant → peptide translation (8-11mers)NetMHCpan-4.1, MHCflurry-2.0
Alternative splicingImmunoPepper, ISOTOPE, ASNEO, SNAFNetMHCpan-4.1, MHCflurry-2.0
Gene fusionsINTEGRATE-neo, AGFusionNetMHCpan-4.1, MHCflurry-2.0
Non-canonicalNeoDisc frameworkNetMHCpan-4.1
MS-validatedDirect immunopeptidomics identificationAlready presented (MS-confirmed)

Prioritization Scoring

MHC-I binding affinity (IC50) Binding rank percentile Immunogenicity (DeepImmuno, PRIME) Tumor-specificity score Clonality Expression level (TPM) MS validation status TCR recognition probability

Outputs

Ranked neoantigen candidate list Per-peptide binding and immunogenicity scores
↓
4
Computational Immunostructural Engineering
Rational peptide design, structural modeling, and self-peptide displacement

Engineering Steps

StepMethodPurpose
Peptide-MHC structural modelingAlphaFold2 / ESMFold / ColabFold3D structure of peptide-MHC complex
TCR-pMHC dockingTCRmodel, ImmuneBuilderModel TCR recognition interface
Binding energy calculationRosetta, FoldXQuantify peptide-MHC and TCR-pMHC binding
Rational peptide engineeringProteinMPNN, computational mutagenesisEnhance MHC binding and TCR recognition
Self-peptide displacementCompetitive binding modelingDisplace antagonistic self-peptides
Immunogenicity optimizationIn silico mutational scanningMaximize TCR activation, maintain tumor specificity

Outputs

Engineered peptide candidates Self-peptide displacement designs Structural models
↓
5
Report Generation
Assemble patient-specific report with visualizations and recommendations

Processing Steps

StepTool(s)

Outputs

Interactive web report (pX_XXXX) PDF report for archival Structured JSON data