Skip to content
LB2

Prostate adenocarcinoma PRAD

Of 60,498 genes, 649 are higher in the tumor than in normal prostate and in whole blood, and are not made by blood immune cells. 241 are protein-coding, and 236 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
495
TCGA primary tumors
Normal tissue
152
52 TCGA tumor-adjacent, 100 GTEx prostate
Whole blood
337
GTEx samples from 328 donors
Survival
495
patients, 92 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

649
candidate genes: 236 with plasma cfRNA evidence, 8 protein-coding without, 405 non-coding
111
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. 0.991
52
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 649 candidates.

Loading candidates…

How the candidates were selected

60,498 genes tested; 1,201 higher in tumor than in normal prostate; 775 also higher than in whole blood; 649 remain after removing genes expressed in blood immune cells; 236 of these have plasma cfRNA evidence, 8 protein-coding candidates have none, and 405 are non-coding genes the plasma sources cannot list.

Figure 2. Genes remaining after each step in prostate 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

Genetoward normaltoward tumorCoef.
  1. RP3-417G15.1+0.48
  2. ZACN+0.46
  3. APOE+0.46
  4. DLX1+0.45
  5. SLIT1+0.45
  6. SNHG19−0.44
  7. LMX1B+0.44
  8. RP11-1042B17.3+0.41
  9. DNAAF3+0.39
  10. FAM166A+0.37
  11. PRAC2+0.37
  12. RP11-356O9.1+0.35
  13. RP11-248G5.9−0.32
  14. LINC01311+0.32
  15. MOV10L1+0.29
  16. RP4-549L20.3+0.29
  17. ARHGEF26+0.28
  18. PCAT1+0.26
  19. TMC5+0.26
  20. RP11-302L19.3+0.25
Show the other 91 genes
  1. HPX−0.24
  2. AC073133.1+0.24
  3. PCAT29+0.24
  4. HSD17B3+0.24
  5. CCDC83−0.23
  6. ATG9B+0.23
  7. SNHG25+0.22
  8. PLAC8L1−0.22
  9. RPRML−0.21
  10. COL10A1+0.21
  11. MKI67+0.21
  12. GDF1+0.20
  13. RP11-621L6.3+0.20
  14. CAMK2N2+0.18
  15. VSTM2L−0.18
  16. VPS9D1-AS1+0.17
  17. TMSB15A−0.17
  18. FLJ20021−0.17
  19. PLEKHH1−0.16
  20. NKX2-3+0.16
  21. CTNND2−0.15
  22. ARHGDIG+0.15
  23. EPHA8+0.15
  24. RP11-618K13.2+0.15
  25. SHISA8+0.15
  26. NUTM2F−0.14
  27. CGREF1+0.14
  28. RP11-589M4.1+0.14
  29. CCDC180+0.14
  30. TOX3−0.14
  31. ZIC5+0.14
  32. HECTD4+0.13
  33. CDK3+0.13
  34. DNAH5+0.13
  35. NEK5+0.13
  36. GAL+0.12
  37. AC012531.25+0.12
  38. MNX1−0.12
  39. ANGPTL3−0.12
  40. PRCAT47+0.11
  41. U73166.2−0.11
  42. RGS11+0.11
  43. COLEC12+0.11
  44. RP11-465N4.4+0.10
  45. SPTY2D1-AS1+0.10
  46. SDK1+0.10
  47. ATP8A2−0.10
  48. THPO+0.09
  49. TMEM184A−0.09
  50. TRPM8+0.08
  51. CTA-246H3.12+0.08
  52. C4A+0.08
  53. CCDC154+0.08
  54. C3orf80+0.08
  55. AP006748.1−0.08
  56. CST2−0.08
  57. AP001610.9+0.08
  58. CERS1+0.07
  59. RP5-984P4.6+0.07
  60. UCN+0.07
  61. FAM64A−0.07
  62. ULBP1+0.07
  63. RP5-1159O4.2+0.06
  64. HOXC6+0.06
  65. PCA3+0.06
  66. QRICH2−0.05
  67. C4B+0.05
  68. HOXC4−0.05
  69. CENPF+0.05
  70. ZNF695−0.04
  71. RP1-228H13.5+0.04
  72. C16orf59−0.04
  73. BIRC7+0.04
  74. ODF3L2+0.04
  75. KCNG3+0.04
  76. TTC6+0.04
  77. OTP+0.03
  78. ANKRD66+0.03
  79. MMP26+0.02
  80. RP11-223A3.1+0.02
  81. LINC00930+0.02
  82. HOXB13+0.02
  83. SPC25+0.01
  84. RP11-680F20.10+0.01
  85. SRCIN1−0.01
  86. CTD-2334D19.1+0.01
  87. UNQ6494+0.01
  88. ADRB1−0.01
  89. STEAP2-AS1−0.00
  90. TRIM67−0.00
  91. CCDC108+0.00
Figure 3. Standardized elastic-net coefficients of the 111 genes selected to separate prostate adenocarcinoma from normal prostate, 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 0.991, 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. CPLX11.78 (1.45–2.19)7.1 × 10−6
  2. KIF4A1.72 (1.38–2.15)4.4 × 10−5
  3. C16orf591.72 (1.41–2.10)7.1 × 10−6
  4. RGS111.71 (1.40–2.09)7.1 × 10−6
  5. FAM64A1.70 (1.38–2.09)2.4 × 10−5
  6. B4GALNT41.68 (1.35–2.10)9.3 × 10−5
  7. RHPN11.63 (1.32–2.01)1.3 × 10−4
  8. PLXNB31.60 (1.28–2.00)3.5 × 10−4
  9. VGF1.60 (1.31–1.95)8.0 × 10−5
  10. ZNF6951.60 (1.27–2.01)5.6 × 10−4
  11. SKA31.56 (1.28–1.90)1.5 × 10−4
  12. IL111.55 (1.27–1.91)3.2 × 10−4
  13. SRCIN11.55 (1.27–1.89)1.9 × 10−4
  14. GABRD1.55 (1.26–1.90)3.9 × 10−4
  15. CBX21.54 (1.26–1.90)3.5 × 10−4
  16. HECTD41.54 (1.26–1.88)2.4 × 10−4
  17. TMEM132A1.54 (1.23–1.91)0.0011
  18. LRRC561.52 (1.22–1.89)0.0014
  19. FEV0.73 (0.62–0.86)0.0014
  20. ADPRHL10.68 (0.55–0.83)0.0014
Show the other 32 genes
  1. ZACN1.72 (1.27–2.33)0.0032
  2. HOXC41.58 (1.21–2.08)0.0044
  3. FAM166A1.53 (1.17–2.02)0.0091
  4. UCN1.49 (1.19–1.86)0.0029
  5. ARHGDIG1.47 (1.19–1.82)0.0029
  6. NOX41.46 (1.17–1.82)0.0044
  7. APOE1.45 (1.18–1.78)0.0029
  8. SLC34A31.45 (1.14–1.85)0.011
  9. EPHA101.44 (1.13–1.82)0.011
  10. HPN-AS11.44 (1.13–1.83)0.011
  11. SLC25A271.43 (1.16–1.76)0.0044
  12. COMP1.43 (1.14–1.78)0.0091
  13. CAMK2N21.42 (1.15–1.76)0.0069
  14. MAPK8IP21.42 (1.16–1.73)0.0034
  15. C2CD4C1.41 (1.13–1.75)0.0089
  16. HSF41.40 (1.13–1.73)0.0086
  17. HPX1.40 (1.16–1.68)0.0032
  18. SNHG251.39 (1.12–1.72)0.010
  19. TPRN1.38 (1.13–1.69)0.0086
  20. RAC31.36 (1.10–1.68)0.016
  21. SAPCD21.35 (1.07–1.69)0.031
  22. LRGUK1.33 (1.08–1.64)0.024
  23. CASKIN11.33 (1.08–1.64)0.024
  24. CERS11.31 (1.07–1.60)0.031
  25. COL10A11.31 (1.05–1.62)0.046
  26. QRICH21.30 (1.07–1.59)0.027
  27. MGAT5B1.30 (1.06–1.60)0.038
  28. ITGBL11.30 (1.05–1.61)0.046
  29. ISX1.28 (1.04–1.56)0.049
  30. GLYATL10.80 (0.66–0.95)0.045
  31. LENG90.78 (0.64–0.95)0.046
  32. SLC13A30.78 (0.65–0.93)0.022
longer PFIshorter PFI
Figure 4. Hazard ratio per standard deviation of tumor expression with 95% confidence interval, log scale, for the 52 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 495 patients (92 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 PRAD

  • 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