Artistic rendering of PD-L1 IHC staining showing tumor and immune cell positivity

PD-L1 Scoring

Automated PD-L1 TPS and CPS Scoring

Checkpoint inhibitor eligibility scoring with multi-assay normalization. Tumor proportion score and combined positive score from a single WSI analysis.

EXAMPLE OUTPUT

PD-L1_result = {
  "assay_clone": "22C3",
  "TPS": 38.4,
  "CPS": 12.1,
  "tumor_cells_pos": 842,
  "immune_cells_pos": 167,
  "confidence": "High"
}

Scoring Systems

TPS, CPS, and IC-Score: The PD-L1 Complexity

PD-L1 IHC is uniquely challenging because the clinical threshold depends on which antibody clone, which scoring system, and which tumor type. Synthia handles all three.

Score Type Definition Cells Counted Common Clinical Use
TPS
Tumor Proportion Score		 Percentage of viable tumor cells with partial or complete membrane staining at any intensity Tumor cells only (excludes immune infiltrate) NSCLC (pembrolizumab); 22C3 and 28-8 assays
CPS
Combined Positive Score		 (PD-L1 positive cells — tumor + lymphocytes + macrophages) ÷ total viable tumor cells × 100 Tumor + immune infiltrate Gastric/GEJ, cervical, head & neck, TNBC; 22C3 assay
IC-Score
Immune Cell Score		 Percentage of tumor area covered by PD-L1 positive immune cells Immune cells within tumor area Urothelial carcinoma; SP142 assay (atezolizumab)

Multi-Assay Normalization: 22C3, 28-8, SP142

One of the most clinically relevant complications of PD-L1 testing is that the three major assays — 22C3 (Dako/Agilent, for Keytruda), 28-8 (Dako/Agilent, for Opdivo), and SP142 (Ventana/Roche, for Tecentriq) — are not interchangeable. They use different antibody epitopes, detection systems, and have different clinical thresholds for the same tumor type.

Synthia's PD-L1 module accepts assay metadata at submission and applies assay-specific calibration curves derived from our pathologist-annotated reference sets. The output report specifies: assay clone, scoring method (TPS or CPS), raw count, and the threshold interpretation for the relevant indication.

Our model separates tumor cell staining from immune cell staining using nuclear and cytoplasmic morphology — critical for CPS calculation accuracy in tumors with heavy immune infiltrate.

NSCLC Thresholds (22C3)

TPS ≥ 50% High expression
TPS 1–49% Low expression
TPS < 1% Negative

Clinical interpretation. Consult oncologist. Synthia output is not a diagnostic report.

Performance Metrics

PD-L1 Internal Validation Results

Internal study, data on file. For investigational use only.

PD-L1
Concordance vs. 3-Pathologist Consensus (22C3 assay)
Internal validation · n=290 WSI · NSCLC biopsies
Metric Value 95% CI Reference Standard
TPS categorical concordance (<1% / 1–49% / ≥50%) 88.3% 84.7–91.6% 3-reader majority vote
Weighted Kappa (TPS categories) 0.88 0.84–0.91 Linear weight
ICC (continuous TPS %) 0.92 0.89–0.94 Pathologist TPS estimate
CPS concordance (vs. manual CPS estimate) 0.89 0.86–0.92 Pearson correlation
TPS ≥50% sensitivity 91.4% 86.8–94.8% Consensus positive (n=105)
TPS ≥50% specificity 94.1% 90.2–96.8% Consensus negative (n=185)
Disclaimer: Internal validation studies. Data on file. Synthia is not FDA cleared. For research and investigational use only. PD-L1 scoring for clinical treatment decisions requires a validated, approved diagnostic assay.

Evaluate PD-L1 Scoring on Your Cases

Submit de-identified NSCLC or gastric cancer WSI with assay metadata for a pilot scoring report.

Request Pilot Access Full Validation Data