Skip to content
LB2

Pancreatic adenocarcinoma PAAD

Of 60,498 genes, 1,136 are higher in the tumor than in normal pancreas and in whole blood, and are not made by blood immune cells. 681 are protein-coding, and 680 of those have supporting plasma Cell-free RNA (cfRNA)RNA fragments that circulate in blood plasma outside cells. Most come from blood cells; a small share comes from other tissues, including tumors. evidence; the rest are mostly non-coding genes the plasma sources cannot check.

Samples compared

Tumor
178
TCGA primary tumors
Normal tissue
171
4 TCGA tumor-adjacent, 167 GTEx pancreas
Whole blood
337
GTEx samples from 328 donors
Survival
177
patients, 104 events (Survival endpointsCurated TCGA outcomes (Liu et al. 2018): OS is overall survival, PFI the progression-free interval and DSS disease-specific survival. LB2 uses whichever has the most events in each cancer type.)

Results in brief

1,136
candidate genes: 680 with plasma cfRNA evidence, 8 protein-coding without, 448 non-coding
27
genes in the tissue panel, Held-out AUCHow well an elastic-net model separates tumor from normal tissue, measured with nested cross-validation on samples the model did not train on. It describes tissue, not the accuracy of a blood test. 0.994
74
genes associated with progression-free interval (FDR < 0.05, 150 tested)

Candidates

Filter by plasma evidence or discrimination, find a gene, and select any point or row for its full record. The table holds all 1,136 candidates.

Loading candidates…

How the candidates were selected

60,498 genes tested; 2,978 higher in tumor than in normal pancreas; 1,423 also higher than in whole blood; 1,136 remain after removing genes expressed in blood immune cells; 680 of these have plasma cfRNA evidence, 8 protein-coding candidates have none, and 448 are non-coding genes the plasma sources cannot list.

Figure 2. Genes remaining after each step in pancreatic adenocarcinoma.

Each comparison uses a two-sided Wilcoxon rank-sum test. A gene passes when its q-value is below 0.05, its log2 fold change is at least 1, and the lower bound of the 95% confidence interval for its AUC is at least 0.70.

Genes made by any of 18 sorted blood immune cell types above 1 nTPM are then removed, because blood cells supply most plasma RNA. Genes missing from that reference, mostly non-coding, are kept.

Plasma evidence only reorders the list. A candidate not yet seen in plasma stays in, because every dataset misses genes. Read the full methods.

Tissue classifier panel

Geneweight toward tumorCoef.
  1. ERICH2+0.40
  2. GOLT1A+0.33
  3. GRPR+0.29
  4. TMEM151A+0.29
  5. AGR2+0.22
  6. UGT1A10+0.18
  7. PHGR1+0.16
  8. ATP6V0A4+0.15
  9. MISP+0.14
  10. FXYD3+0.13
  11. ALPI+0.12
  12. CIDEC+0.12
  13. FAM83A+0.11
  14. B3GNT3+0.11
  15. C2orf70+0.09
  16. IRX5+0.08
  17. FOXA2+0.08
  18. JPH1+0.07
  19. MARCH4+0.06
  20. PRAP1+0.05
Show the other 7 genes
  1. SERPINB5+0.05
  2. PHYHIPL+0.04
  3. MYEOV+0.04
  4. PITX1+0.03
  5. ABCA12+0.01
  6. HAPLN1+0.01
  7. SPDEF+0.00
Figure 3. Standardized elastic-net coefficients of the 27 genes selected to separate pancreatic adenocarcinoma from normal pancreas, largest first.

An elastic-net logistic regression was trained on the 300 top-ranked candidates to tell tumor from normal tissue. Its Held-out AUCHow well an elastic-net model separates tumor from normal tissue, measured with nested cross-validation on samples the model did not train on. It describes tissue, not the accuracy of a blood test. is 0.994, from nested five-fold cross-validation.

Before training, the TCGA-versus-GTEx offset was removed from the expression matrix with the tumor/normal contrast protected, as in the tissue comparison. That correction is fitted once on all samples, so the held-out AUC is not fully independent of it.

This number describes tumor and normal tissue, where separation is expected to be near perfect. It is not the accuracy of a blood test: that has to be measured in plasma from patients and controls.

Association with survival

Gene0.250.5124HR per SD (95% CI)q-value
  1. SERPINB51.91 (1.42–2.57)0.0014
  2. GRHL21.86 (1.32–2.60)0.0024
  3. KRT191.86 (1.39–2.49)0.0018
  4. FERMT11.80 (1.33–2.42)0.0018
  5. SCEL1.78 (1.37–2.31)0.0014
  6. MYEOV1.70 (1.31–2.20)0.0018
  7. EPN31.63 (1.25–2.11)0.0024
  8. UGT1A101.60 (1.25–2.04)0.0022
  9. S100A141.59 (1.24–2.04)0.0022
  10. EVPL1.58 (1.24–2.01)0.0022
  11. MAL21.56 (1.23–1.97)0.0022
  12. LAMA31.53 (1.25–1.89)0.0018
  13. PSCA1.53 (1.23–1.89)0.0018
  14. PKP31.52 (1.21–1.91)0.0024
  15. PCDH11.52 (1.23–1.88)0.0018
  16. FAM83H1.49 (1.21–1.85)0.0022
  17. FOXL11.47 (1.21–1.79)0.0018
  18. TMEM151A0.70 (0.58–0.85)0.0022
  19. MSI10.69 (0.57–0.84)0.0022
  20. SEZ6L20.68 (0.56–0.83)0.0018
Show the other 54 genes
  1. PRSS81.69 (1.25–2.28)0.0037
  2. C6orf1321.60 (1.23–2.09)0.0031
  3. C1orf1061.59 (1.21–2.10)0.0042
  4. LIPH1.56 (1.21–2.02)0.0037
  5. KLK111.53 (1.20–1.94)0.0034
  6. LAD11.52 (1.20–1.93)0.0031
  7. B3GNT31.50 (1.17–1.93)0.0069
  8. CLDN41.49 (1.19–1.88)0.0034
  9. FUT31.49 (1.16–1.92)0.0070
  10. SYT81.48 (1.18–1.86)0.0037
  11. ESRP11.48 (1.17–1.87)0.0052
  12. OVOL11.47 (1.18–1.83)0.0037
  13. WNT7B1.46 (1.15–1.86)0.0074
  14. FOXQ11.46 (1.17–1.82)0.0042
  15. MST1R1.46 (1.18–1.80)0.0031
  16. PITX11.46 (1.15–1.85)0.0074
  17. XDH1.43 (1.14–1.78)0.0069
  18. STEAP11.40 (1.13–1.74)0.0076
  19. KRT151.40 (1.15–1.71)0.0047
  20. ERN21.39 (1.11–1.74)0.013
  21. PQLC2L1.39 (1.09–1.75)0.019
  22. ABCA121.38 (1.12–1.71)0.0091
  23. C1orf1161.38 (1.09–1.74)0.020
  24. SDC11.37 (1.12–1.67)0.0074
  25. MISP1.37 (1.09–1.71)0.017
  26. AQP51.37 (1.11–1.68)0.011
  27. CCDC64B1.36 (1.07–1.72)0.027
  28. SPDEF1.35 (1.13–1.62)0.0051
  29. PRR151.35 (1.10–1.65)0.012
  30. IRF61.32 (1.04–1.68)0.042
  31. ABHD17C1.31 (1.08–1.59)0.017
  32. FA2H1.31 (1.08–1.58)0.017
  33. PLA2G101.31 (1.06–1.61)0.027
  34. FUT21.31 (1.07–1.59)0.021
  35. BAIAP2L11.31 (1.07–1.59)0.020
  36. MYH141.30 (1.06–1.59)0.027
  37. AGR21.30 (1.06–1.59)0.027
  38. CGN1.30 (1.07–1.58)0.024
  39. ESRP21.29 (1.05–1.58)0.034
  40. FXYD31.28 (1.04–1.58)0.041
  41. CEACAM51.28 (1.06–1.55)0.027
  42. SLC15A11.28 (1.03–1.58)0.047
  43. ELF31.27 (1.04–1.55)0.040
  44. MYOM31.26 (1.05–1.51)0.029
  45. EXPH51.25 (1.04–1.52)0.041
  46. C6orf2231.25 (1.04–1.52)0.041
  47. FOXA20.82 (0.70–0.96)0.035
  48. RAB170.81 (0.69–0.94)0.020
  49. MARCH40.79 (0.66–0.94)0.026
  50. TACC20.78 (0.64–0.94)0.027
  51. C2orf720.77 (0.64–0.93)0.019
  52. ASPHD10.77 (0.64–0.93)0.017
  53. PHYHIPL0.77 (0.64–0.92)0.015
  54. ARFGEF30.73 (0.61–0.88)0.0047
longer PFIshorter PFI
Figure 4. Hazard ratio per standard deviation of tumor expression with 95% confidence interval, log scale, for the 74 genes with FDR < 0.05.

For the 150 top-ranked candidates, a Cox model relates tumor expression, as a continuous value, to progression-free interval in 177 patients (104 events). No high/low cutpoint is searched for, since optimized cutpoints inflate false positives.

A Hazard ratio per SDFrom a Cox model with the gene’s tumor expression as a continuous variable: the change in hazard for each one-standard-deviation increase. Above 1, higher expression goes with a shorter time to the event; below 1, with a longer one. above 1 means higher expression goes with a shorter progression-free interval. These are associations in tissue, not evidence that a blood level predicts outcome.

Caveats for PAAD

  • Only 4 TCGA adjacent samples anchor the correction for study differences between TCGA and GTEx normals.
  • The blood comparison sets TCGA tumors against GTEx blood. Study and biology cannot be separated there, which is why genes made by blood immune cells are removed as well.

Limitations that apply to every cancer type