Supplementary Materials

The PDF file includes:

  • Fig. S1. Biochemical features of caTCRs showing significant differences from non-cancer TCRs.
  • Fig. S2. CDR3 length distribution and prediction accuracy of CNN models for each CDR3 length.
  • Fig. S3. DeepCAT prediction is independent of HLA allele types.
  • Fig. S4. Distribution of raw DeepCAT outputs for patients with cancer and healthy donors.
  • Fig. S5. Cancer score estimation is independent of library size or patient age.
  • Fig. S6. Ninety-five percent confidence interval plots for ROC curves and waterfall plots for the cohorts of patients with cancer in Fig. 3.
  • Fig. S7. ROC analysis for all public PBMC cohorts with both treated and treatment-na├»ve patients.
  • Fig. S8. Variation of cancer scores in the time course cohort.
  • Fig. S9. CDR3 length distribution of healthy individuals and patients with cancer.
  • Fig. S10. ROC curves for cancer scores predicted using four different combinations.
  • Fig. S11. Evaluation of prediction accuracy for early-stage cancers in two recent studies.
  • Fig. S12. Boxplots showing the positive and negative prediction values at different cancer score cutoffs.
  • Fig. S13. Cancer score association with library size or age using the RNA-based samples.
  • Fig. S14. Cancer score distributions of patients with non-cancer chronic conditions compared against healthy controls.
  • Table S1. Summary information of caTCRs obtained from TCGA data.
  • Table S2. Summary of datasets used in this study.
  • Table S3. Clinical information for TCR-seq samples generated in this study.

[Download PDF]

Other Supplementary Material for this manuscript includes the following:

  • Data file S1 (Microsoft Excel format). DeepCAT estimation of cancer scores across all the sample cohorts.