Supplementary information to “IPI59: an actionable biomarker to improve treatment response in serous ovarian carcinoma patients”

1. Package/annotation versions

Unless noted otherwise, calculations are carried out in R [1], version 3.2.2. The versions for each package considered in the manuscript are as follow: KEGG.db_3.2.2, glmnet_2.0-2, SGL_1.1, survival_2.38-3, affy_1.48.0, hgu133plus2.db_3.2.2, hthgu133a.db_3.2.2, and curatedOvarianDate_1.3.5.

2. TCGA data

The TCGA data was downloaded from the TCGA data portal at http://tcga.cancer.gov on 08/25/2015. Gene expression files were downloaded before normalization (level1, .CEL).

Clinical data definition. In the clinical data file, 586 patients were available. The ‘vital_status’ variable was used as vital status, the ‘death_to_days’ and the ‘last_contact_days_to’ variables were used for survival and follow-up time. We converted the time from days to months (30 days per month) in our analysis. For cytoreduction status, the ‘residual_disease_largest_nodule’ column was used. Cytoreduction status of any residual disease (RD+) was defined as having a residual tumor size of 1-10 mm, 11-20 mm, or >20 mm; and no macroscopic disease was defined as no residual disease (RD-), consistent with the most recent definitions [2–4] and supported by Supplementary Figures S1 and S2. The follow-up time to new tumor event after the initial treatment variable was defined by either ‘days_to_new_tumor_event_after_initial_treatment’ or ‘days_to_new_tumor_event_additional_surgery_ procedure’ in clinical data.

Treatment data definition. 507 patients have information regarding adjuvant treatments in the drug file. Treatments that are given to patients after recurrence were not considered for patient grouping (adjuvant therapy vs. enhanced adjuvant therapy). The ‘pharmaceutical_tx_started_days_to’ and ‘pharmaceutical_tx_ended_days_to’ variables were used to derive treatment start and end dates. Table S2 shows a summary of TCGA patient treatments.

Excluded patients. The clinical data and the gene expression data share 520 patients. Within 520 patients, we exclude one patient with missing vital status (TCGA-04-1519) and three patients with missing survival/follow-up times (TCGA-24-0968, TCGA-04-1357, and TCGA-04-1360). For the biomarker identification, 62 patients with low tumor stage (stage I and stage II) or tumor grade 1 were excluded. Among the 507 patients with clinical data, 411 were included for the analysis of association of IPI59 with treatment due to availability of gene expression data and other clinical variables. Twenty-five patients who were noted as receiving adjuvant therapy were also excluded since the treatment start date was longer than 8 months post surgery.

Patients having biopsies at recurrence. There were 16 patients for which expression profiles were available from both primary and recurrent tumors (Table S1). Primary tumor arrays were used for extended pathway-index modeling; recurrent tumor expression for 13 patients was used to investigate changes in IPI59 since three patients (TCGA-13-1819, TCGA-29-1707, TCGA-61-2008) were removed from the analysis due to low tumor stage (Stage II).

3. Validation data

Clinical data definition. All the clinical variables are used as they defined in the paper. For cytoreduction status, we followed the recent distinction of RD- vs RD+ in GSE9891. For other studies (GSE19829, GSE63885, GSE30161, GSE26712, and GSE32062) that this information was not available, we used the provided status label; the patients with optimal cytoreduction are included in the RD- group and the patients with sub-optimal are included in RD+ group.

Excluded patients. Within total of 917 patients, we exclude 33 patients with missing survival time. Among the remaining patients, 59 patients with low tumor stage (stage I and stage II) or tumor grade 1 were excluded.

Figures

Figure S1: Kaplan-Meier curves of overall survival for 458 TCGA patients as a function of cytoreduction status (left). In the middle panel, the survival predictions are shown for the traditional definition of optimal (blue) and sub-optimal (red), and the survival predictions for RD- (blue) and RD+ (red) are shown in the right.

Figure S2: Kaplan-Meier curves of progression-free survival for 458 TCGA patients as a function of cytoreduction status (left). In the middle panel, the survival predictions are shown for the traditional definition of optimal (blue) and sub-optimal (red), and the survival predictions for RD- (blue) and RD+ (red) are shown in the right.

Table S1 The primary solid tumor arrays (01) were used for the integrative pathway index (IPI) modeling. The recurrent solid tumor arrays (02) were used for paired sample analysis (PSA).

Connection - Clinical / Connection – Expression / Time to Rec. (day) / Use for IPI / Use for PSA
TCGA-13-0791
TCGA-13-0913
TCGA-13-1489
TCGA-
13-1817
TCGA-13-1819
TCGA-23-1023
TCGA-24-1852
TCGA-29-1692
TCGA-29-1705
TCGA-29-1707
TCGA-29-1710
TCGA-29-1770
TCGA-29-2414
TCGA-61-1916
TCGA-61-2008
TCGA-61-2095 / 02
01
02
01
02
01
01
02
01
02
02
01
01
02
01
02
01
02
01
02
01
02
01
02
01
02
01
02
02
01
02
01 / TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E01_586136
BONES_p_TCGA_Batch8_9_RNA_HT_HG.U133A_96.HTA_E01_297940
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E04_586142
PSHAW_p_TCGA_Batch11_RNA_HT_HG.U133A_96.HTA_A09_433742
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E07_586164
SEDUM_p_TCGA_Batch14_Expr_HT_HG.U133A_96.HTA_C04_458686
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E10_586154
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_F01_586114
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_F04_586116
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_F07_586110
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_A06_586078
AGARS_p_TCGA_B12_RNA_ReDo_HT_HG.U133A_96.HTA_C07_443130
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E02_586138
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E05_586160
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E08_586166
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E11_586156
LENOS_p_TCGA_Batch18and19_HT_HG.U133A_96.HTA_B10_491120
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_F08_586108
LENOS_p_TCGA_Batch18and19_HT_HG.U133A_96.HTA_C04_491124
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_C05_586182
LENOS_p_TCGA_Batch18and19_HT_HG.U133A_96.HTA_C10_491142
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E03_586140
LENOS_p_TCGA_Batch18and19_HT_HG.U133A_96.HTA_D11_491238
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E06_586162
DITTY_p_TCGA_B_24_AMplate1_2_HT_HG.U133A_96.HTA_B11_547252
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E09_586152
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_E12_586158
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_F03_586118
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_F06_586104
DURUM_p_TCGA_B22_23_Expr_HT_HG.U133A_96.HTA_B04_543644
TARRE_p_MultiPlate_TCGA_SS_MA_Ref_HT_HG.U133A_96.HTA_D12_586150
DURUM_p_TCGA_B22_23_Expr_HT_HG.U133A_96.HTA_C05_543710 / 281
871
806
541
429
490
1009
544
283
1201
315
377
552
326
818
445 / X
O
X
O
X
O
O
X
O
X
X
O
O
X
O
X
O
X
O
X
O
X
O
X
O
X
O
X
X
O
X
O / O
O
O
O
O
O
O
O
X
X
O
O
O
O
O
O
O
O
X
X
O
O
O
O
O
O
O
O
X
X
O
O
Difference between multiple arrarys are caused by “Sample” and “Vial”
(See the right figure from TCGA website)
Sample: 01 – Primary solid tumor
Sample: 02 – Recurrent solid tumor /

Table S2: Summary of TCGA patient treatment.

Treatment Group / Treatments / Dictated names / Number of patients*
Adjuvant therapy / Platinum line / Carboplatin, carboplatin, carbopaltin, Carbobplatin, Carbo, Carboplatin 6th, Carboplatinum, Carbplatin Ciplastin, Cisplatin, Cisplatinum, CCDP / 573
Taxane line / Paciltaxal, Paciltaxel, Pacitaxel, Pacliatxel, Pacliltaxel, paclitaxel, Paclitaxel, Paclitaxel; Albumin-Bount, Paclitaxil, taxol, Taxol, Taxol/Carboplatin, Taxol or Taxotere, Taxotere, Taxoterecin, Docetaxel, Doxetaxel / 544
Enhanced adjuvant therapy / Gemcitabine / gemcitabine, Gemcitabine, Gemcitabine HCI, Gemcitibine, Gemicitabine, Gemzar, Gemzet / 39
Bevacizumab / Bevacizumab, Avastin / 13
Topotecan / Topetecan, Topotecan, Topotecan/Carboplatin, Toptecan / 12
Extra / Doxil, Doxorubicin, Adriamycin, Amitostin, Amitostine, Tamoxifen, Vincristine, Oxaliplatin, Cetuximab, Etoposide, Femara, Cytoxan, GOG182, GOG 218, Herceptin, Hexalin, Interferon gamma, Investigal drug Avastin, Megace, Melphalan, Mitoxantrone, Ovarex MA6 B43.13 / 62
Remove / Abagovomab or Placebo, Abagovomab vs Placebo, Abagovomag vs Placebo, Avastin/bevacizumab/placebo, Bevacizamab/Placebo,
Bevacizumab or placebo, Bevacizumab or Placebo, Bevacizumab vs Placebo, Bevacizumiab/ versus Placebo, Ovarex/Placebo
Decodron, [Not Available] / 16

* Patients are listed more than once when they received multiple treatments

3. Results by individual pathways

Figure S3: The upper left panel is the same as the left panel of Figure 1 (OS), but calculated using the index constructed from 9 probes (9 genes) selected from the Axon Guidance pathway. The upper right panel is the same as the left panel of Figure 2 (PFS). The bottom panels correspond to Figure 4, again for the Axon Guidance pathway.

Figure S4: The upper left panel is the same as the left panel of Figure 1 (OS), but calculated using the index constructed from 18 probes (18 genes) selected from the Neuroactive Ligand-Receptor Interaction pathway. The upper right panel is the same as the left panel of Figure 2 (PFS). The bottom panels correspond to Figure 4, again for the Neuroactive Ligand-Receptor Interaction pathway.

Figure S5: The upper left panel is the same as the left panel of Figure 1 (OS), but calculated using the index constructed from 12 probes (11 genes) selected from the Neurotrophin Signaling pathway. The upper right panel is the same as the left panel of Figure 2 (PFS). The bottom panels correspond to Figure 4, again for the Neurotrophin Signaling pathway.

Figure S6: The upper left panel is the same as the left panel of Figure 1 (OS), but calculated using the index constructed from 15 probes (14 genes) selected by the pathway-index model applied to the GnRH Signaling pathway. The upper right panel is the same as the left panel of Figure 2 (PFS). The bottom panels correspond to Figure 4, again for the GnRH Signaling pathway.

Figure S7: The upper left panel is the same as the left panel of Figure 1 (OS), but calculated using the index constructed from 21 probes (19 genes) selected from the Wnt Signaling pathway. The upper right panel is the same as the left panel of Figure 2 (PFS). The bottom panels correspond to Figure 4, again for the Wnt Signaling pathway.

Figure S8: The upper left panel is the same as the left panel of Figure 1 (OS), but calculated using the index constructed from 21 probes (20 genes) selected from the Osteoclast Differentiation pathway. The upper right panel is the same as the left panel of Figure 2 (PFS). The bottom panels correspond to Figure 4, again for the Osteoclast Differentiation pathway.

4. Top 6 pathways identified by the pathway-index model and Group lasso

Table S4: Top 6 pathways with selected genes after applying the pathway-index model and Group lasso (total of 81 probes). The second column shows p-value from log-rank test between patient groups defined by prognosis (high-risk vs. low-risk). The third column shows the number of nonzero coefficients over pathway size. Selected 59 genes after applying the integrative pathway-index model are shown with bold face (IPI59).

Pathway name / Predicted score (p-value) / Nonzero coeff/Pathway size / Direction to survival / Gene Symbol
Axon guidance / 2.08e-12 / 9/209 / Good / EPHA1, SEMA4G, SEMA4A
Poor / CDC42, GNAI1, PLXNB1, PPP3R1, RAC1, UNC5B
Neuroactive ligand-receptor interaction / 2.40e-10 / 18/376 / Good / AVPR1A, AVPR1B, CHRNA1, CHRNA4, CNR2, EDNRB, GABRG2, GRIN1, GRM7, THRA, VIPR2, LPAR2, GABBR2
Poor / ADRA1A, ADRA2A, PTGER3, PTGFR, TSHR
Neurotrophin / 2.78e-10 / 12/229 / Good / MAP3K3, MAPK11, PSEN2
Poor / AKT2, CDC42, MAP2K1, PTPN11, RAC1, RPS6KA2, YWHAB (217717_s_at) , YWHAB (217718_s_at), YWHAZ
GnRH signaling / 8.31e-11 / 15/192 / Good / ADCY2, ATF4, ELK1 (203617_x_at), ELK1 (210850_s_at), MAP3K3, PLA2G4A, MAPK11, PLA2G2D, PLA2G12A
Poor / CACNA1C, CDC42, EGFR, JUN, KRAS, MAP2K1
Wnt signaling pathway / 3.33e-16 / 21/261 / Good / DVL3, PPP2R1B, PPP2R5D, WNT5A, FZD5, FZD10, DAAM2
Poor / APC (203525_s_at), APC(203526_s_at), JUN, SMAD3, NFATC1, NFATC3, PPP2R1B, PPP3CA, PPP3R1, RAC1 (208641_s_at), RAC1 (208640_s_at), DKK4, PORCN, WNT5B
Osteoclast differentiation / 5.19e-13 / 21/244 / Good / ITGB3, LCK, TNFRSF11B, MAPK11, TGFB2, TGFB2, GAB2, AKT3, LILRB1
Poor / AKT2, FOSB, JUNB, NFATC1, PPP3CA, MAP2K1, RAC1 (208641_s_at), RAC1 (208640_s_at), STAT2, TGFBR2, TNFSF11, LILRA2

* Number of genes do not sum to 59 due to genes that show up in more than one pathway.

Reference

1. R Development Core Team R (2014) R: A Language and Environment for Statistical Computing. R Found Stat Comput 1:409. doi: 10.1007/978-3-540-74686-7

2. Tucker SL, Gharpure K, Herbrich SM, et al. (2014) Molecular biomarkers of residual disease after surgical debulking of high-grade serous ovarian cancer. Clin Cancer Res 20:3280–8. doi: 10.1158/1078-0432.CCR-14-0445

3. Chang S-J, Bristow RE, Ryu H-S (2012) Impact of complete cytoreduction leaving no gross residual disease associated with radical cytoreductive surgical procedures on survival in advanced ovarian cancer. Ann Surg Oncol 19:4059–67. doi: 10.1245/s10434-012-2446-8

4. du Bois A, Reuss A, Pujade-Lauraine E, et al. (2009) Role of surgical outcome as prognostic factor in advanced epithelial ovarian cancer: a combined exploratory analysis of 3 prospectively randomized phase 3 multicenter trials: by the Arbeitsgemeinschaft Gynaekologische Onkologie Studiengruppe Ovarialkarzin. Cancer 115:1234–44. doi: 10.1002/cncr.24149