immunology.sciencemag.org/cgi/content/full/5/44/eaaz3199/DC1 Supplementary Materials for Tumor neoantigenicity assessment with CSiN score incorporates clonality and immunogenicity to predict immunotherapy outcomes Tianshi Lu, Shidan Wang, Lin Xu, Qinbo Zhou, Nirmish Singla, Jianjun Gao, Subrata Manna, Laurentiu Pop, Zhiqun Xie, Mingyi Chen, Jason J. Luke, James Brugarolas, Raquibul Hannan, Tao Wang* *Corresponding author. Email: [email protected]Published 21 February 2020, Sci. Immunol. 5, eaaz3199 (2020) DOI: 10.1126/sciimmunol.aaz3199 The PDF file includes: Materials Fig. S1. Predictive power of neoantigen load. Fig. S2. Predictive power of the neoantigen fitness model. Fig. S3. Association of CSiN (A), neoantigen loads (B), and neoantigen fitness (C) with IL- 2/SAbR treatment response in ccRCC patients. Fig. S4. Prognostic power of neoantigen load. Fig. S5. Prognostic power of the neoantigen fitness model. Fig. S6. Association of CSiN, neoantigen loads, and neoantigen fitness with prognosis of patients with pediatric ALL and patients with LIHC. Fig. S7. Cartoon showing the workflow of calculation of CSiN scores. Fig. S8. The CSiN plot for the primary tumor of XP397 from the UTSW KCP cohort is shown. Fig. S9. Validity of neoantigen predictions. Fig. S10. The average number of neoantigens generated by each type of mutations. Fig. S11. Demonstrating independence of CSiN from mutation load/neoantigen and transcriptomic-based biomarkers. Fig. S12. Association of CSiN with metastasis. Fig. S13. Validating the predictive power of CSiN, neoantigen load and neoantigen fitness model using OS/PFS as the criterion. Fig. S14. Assessing the intra-tumor heterogeneity of CSiN and neoantigen loads. Fig. S15. Calculating CSiN with only exome-seq data. Fig. S16. Using the median + 2 x interquartile range cutoff on neoantigen load. Fig. S17. Predictive value of class I-specific CSiN and class II-specific CSiN. Fig. S18. Predictive value of class I-specific neoantigen fitness model measured by survival analyses (limiting to 9-mers from missense mutations). Fig. S19. Predictive value of class I-specific neoantigen fitness model measured by categorical response variables (limiting to 9-mers from missense mutations). Fig. S20. Predictive value of CSiN for the patients with high T eff signature expression.
31
Embed
Supplementary Materials for - Science...Fig. S3. Association of CSiN (A), neoantigen loads (B), and neoantigen fitness (C) with IL-2/SAbR treatment response in ccRCC patients. 3 patients
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Published 21 February 2020, Sci. Immunol. 5, eaaz3199 (2020)
DOI: 10.1126/sciimmunol.aaz3199
The PDF file includes:
Materials Fig. S1. Predictive power of neoantigen load. Fig. S2. Predictive power of the neoantigen fitness model. Fig. S3. Association of CSiN (A), neoantigen loads (B), and neoantigen fitness (C) with IL-2/SAbR treatment response in ccRCC patients. Fig. S4. Prognostic power of neoantigen load. Fig. S5. Prognostic power of the neoantigen fitness model. Fig. S6. Association of CSiN, neoantigen loads, and neoantigen fitness with prognosis of patients with pediatric ALL and patients with LIHC. Fig. S7. Cartoon showing the workflow of calculation of CSiN scores. Fig. S8. The CSiN plot for the primary tumor of XP397 from the UTSW KCP cohort is shown. Fig. S9. Validity of neoantigen predictions. Fig. S10. The average number of neoantigens generated by each type of mutations. Fig. S11. Demonstrating independence of CSiN from mutation load/neoantigen and transcriptomic-based biomarkers. Fig. S12. Association of CSiN with metastasis. Fig. S13. Validating the predictive power of CSiN, neoantigen load and neoantigen fitness model using OS/PFS as the criterion. Fig. S14. Assessing the intra-tumor heterogeneity of CSiN and neoantigen loads. Fig. S15. Calculating CSiN with only exome-seq data. Fig. S16. Using the median + 2 x interquartile range cutoff on neoantigen load. Fig. S17. Predictive value of class I-specific CSiN and class II-specific CSiN. Fig. S18. Predictive value of class I-specific neoantigen fitness model measured by survival analyses (limiting to 9-mers from missense mutations). Fig. S19. Predictive value of class I-specific neoantigen fitness model measured by categorical response variables (limiting to 9-mers from missense mutations). Fig. S20. Predictive value of CSiN for the patients with high Teff signature expression.
Fig. S21. Predictive value of CSiN for the patients treated by sunitinib and by atezolizumab in the IMmotion150 cohort. Fig. S22. Predictive value of CSiN for all the patients in the Hellmann cohort. Fig. S23. Boxplots showing distribution of CSiN scores in quartiles of tumor clone number determined by pyclone. Table S1. The patient cohorts used in this study. Table S3. P values and false discovery rates of the tested cohorts shown in Figs. 2 and 3.
Other Supplementary Material for this manuscript includes the following: (available at immunology.sciencemag.org/cgi/content/full/5/44/eaaz3199/DC1)
Table S2. Processed mutation, expression and neoantigen data of the IL-2 cohort (in Excel spreadsheet). Data file S1. Raw data file for Figs. 1 to 3 (in Excel spreadsheet).
Materials
Fig. S1. Predictive power of neoantigen load. The analyses are the same as in Fig. 2, except
that neoantigen loads are considered.
Fig. S2. Predictive power of the neoantigen fitness model. The analyses are the same as in Fig.
2, except that neoantigen fitness model is considered.
Fig. S3. Association of CSiN (A), neoantigen loads (B), and neoantigen fitness (C) with IL-
2/SAbR treatment response in ccRCC patients. 3 patients with complete response (CR), 1
patient with partial response (PR), and 2 patients with stable disease (SD) for more than 6
months form the DCB group. 3 patients with stable disease (SD) less than 6 months and 7
patients with progressive disease (PD) form the NCB group.
Fig. S4. Prognostic power of neoantigen load. The analyses are the same as in Fig. 3, except
that neoantigen loads are considered.
Fig. S5. Prognostic power of the neoantigen fitness model. The analyses are the same as in Fig.
3, except that neoantigen fitness model is considered.
Fig. S6. Association of CSiN, neoantigen loads, and neoantigen fitness with prognosis of
patients with pediatric ALL and patients with LIHC. P values for logrank tests are shown.
(A-C) 103 pediatric and young adult T-lineage acute lymphoblastic leukemia patients were
analyzed. (D-F) 292 TCGA LIHC patients were analyzed. The top 40 LIHC patients were
designated as having “High T cells”, as LIHC is less immunogenic than the other tumor types
investigated in this study.
Fig. S7. Cartoon showing the workflow of calculation of CSiN scores.
Whole Exome Sequencing
RNA Sequencing
Somatic mutations: SNPs, Indels, stoploss
mutationsHLA Typing
Predict HLA-peptide binding affinity
Identify neoantigens with high binding affinity and with expression
level >1 RPKM
Candidate neoantigens
For each mutation
VafiVaf
Vaf is normalized by average vaf of all
mutations in the patientneoantigen loadi
neoantigen load
Number of neoantigens associated with each
mutations; normalized by the
average per mutation neoantigen load across all
mutations
Fundamental building block of CSiNVafi neoantigen loadi
Vaf neoantigen loadx
Under a binding strength cutoff Ck:Vafi neoantigen loadi
Vaf neoantigen load
Under k binding strength cutoffs C1….Ck:
(C1+C2…+Ck)
k
i=1…n
x∑1
ICk=log( )
CSiN =
A percentile rank cutoff Ck is set so that neoantigens with HLA
binding affinity stronger than Ck
are convolved for calculation
CSiN is calculated by the average of the products calculated with the k
cutoffs on binding affinity
Table S1. The patient cohorts used in this study.
Cohort ID Disease type Immunotherapy treatment Raw data Total # patients