Skip to content
LB2

Kidney renal clear cell carcinoma KIRC

Of 60,498 genes, 1,267 are higher in the tumor than in normal kidney and in whole blood, and are not made by blood immune cells. 488 are protein-coding, and 491 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
530
TCGA primary tumors
Normal tissue
100
72 TCGA tumor-adjacent, 28 GTEx kidney
Whole blood
337
GTEx samples from 328 donors
Survival
528
patients, 173 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

1,267
candidate genes: 491 with plasma cfRNA evidence, 2 protein-coding without, 774 non-coding
169
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
59
genes associated with overall survival (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 1,267 candidates.

Loading candidates…

How the candidates were selected

60,498 genes tested; 3,311 higher in tumor than in normal kidney; 1,772 also higher than in whole blood; 1,267 remain after removing genes expressed in blood immune cells; 491 of these have plasma cfRNA evidence, 2 protein-coding candidates have none, and 774 are non-coding genes the plasma sources cannot list.

Figure 2. Genes remaining after each step in kidney renal clear cell carcinoma.

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. RP11-547D24.1+0.26
  2. GABRD+0.25
  3. AHNAK2+0.20
  4. SCN1B+0.20
  5. FATE1+0.19
  6. HSF4+0.18
  7. ANO4+0.17
  8. NEIL3+0.17
  9. RAB42+0.17
  10. CA9+0.17
  11. ALPK2−0.16
  12. GRIK3+0.16
  13. TMEM74B+0.16
  14. IRX3−0.16
  15. PREX2−0.16
  16. LINGO1−0.15
  17. HS3ST2+0.15
  18. HSPB8+0.15
  19. ADRA1B−0.15
  20. ELOVL2+0.14
Show the other 149 genes
  1. CDCA2+0.14
  2. NDUFA4L2+0.14
  3. TMEM233+0.14
  4. BIRC7+0.13
  5. EDA2R+0.13
  6. HTRA4+0.13
  7. NRG3+0.13
  8. NRARP−0.13
  9. DLX1+0.13
  10. NOVA2−0.13
  11. TMEM155+0.12
  12. CFAP74+0.12
  13. EPHA3−0.12
  14. FOXF1−0.12
  15. VWF+0.11
  16. LAMA4+0.11
  17. SOX11+0.11
  18. NNMT+0.11
  19. FABP7+0.11
  20. RASL12−0.10
  21. CCL18+0.10
  22. DOC2A+0.10
  23. NKAIN1+0.09
  24. USH1C−0.09
  25. TMEM133+0.09
  26. STC2+0.09
  27. INHBB+0.09
  28. TBX15+0.09
  29. TRPA1−0.09
  30. C4orf47−0.08
  31. KCNK3−0.08
  32. SIX1+0.08
  33. ASPHD1+0.08
  34. FJX1+0.08
  35. ATP2B2+0.08
  36. EXOC3L1−0.08
  37. CTAGE9+0.07
  38. AVPR1B−0.07
  39. NFE2L3−0.07
  40. MLIP+0.07
  41. MYEOV+0.07
  42. FABP6+0.07
  43. CCDC74A−0.07
  44. CP−0.07
  45. PCDH17−0.07
  46. DCLK3+0.07
  47. COL23A1+0.07
  48. RASD2+0.07
  49. PSORS1C1+0.07
  50. C5orf46+0.06
  51. PRR16−0.06
  52. PPP1R3G+0.06
  53. GPR4−0.06
  54. TMEM200B−0.06
  55. CREB3L3−0.06
  56. JAG2−0.06
  57. FLT1−0.06
  58. TP73+0.06
  59. CPA6+0.06
  60. SLC28A1−0.05
  61. CYP3A5+0.05
  62. ACE−0.05
  63. BCO1−0.05
  64. CDH8−0.05
  65. ADAMTS7+0.05
  66. HMCN1−0.05
  67. GLIS1−0.05
  68. ASPM+0.05
  69. COL5A3+0.05
  70. HEYL−0.05
  71. SCGN+0.05
  72. STX1B−0.05
  73. TNFSF9+0.04
  74. HSPG2−0.04
  75. FAM64A+0.04
  76. NETO2+0.04
  77. KIAA0895L+0.04
  78. PLXNA3−0.04
  79. SKA3+0.04
  80. C9orf172+0.04
  81. HLA-G+0.04
  82. PPFIA4−0.04
  83. GDF6+0.04
  84. OSMR−0.04
  85. C20orf202−0.04
  86. OPN4+0.04
  87. DLX5+0.04
  88. C3orf67−0.04
  89. KISS1R+0.04
  90. FOXM1+0.03
  91. P4HA2−0.03
  92. MYO3A−0.03
  93. HAPLN4−0.03
  94. C1QL1−0.03
  95. ARHGEF39+0.03
  96. KRT36+0.03
  97. FAM163A+0.03
  98. FFAR4+0.03
  99. PERM1−0.03
  100. FKBP10+0.03
  101. NPTX2+0.03
  102. BRIP1+0.03
  103. NOTCH4−0.03
  104. SHROOM1−0.03
  105. SLCO1C1−0.03
  106. GAREML+0.03
  107. PROS1−0.03
  108. PNCK+0.02
  109. SLCO2B1−0.02
  110. CDON−0.02
  111. FN1−0.02
  112. EN1+0.02
  113. PADI1+0.02
  114. SLC17A2+0.02
  115. EXOC3L4−0.02
  116. CENPF+0.02
  117. GABRE−0.02
  118. ANO1−0.02
  119. LRRC75B−0.02
  120. CDK18−0.02
  121. PIEZO2+0.02
  122. MASP1+0.02
  123. COX4I2+0.02
  124. CXCL11+0.01
  125. COL27A1−0.01
  126. NCAPG+0.01
  127. ADAM18+0.01
  128. SLC10A6−0.01
  129. PCDHGC5−0.01
  130. COL5A1−0.01
  131. PRUNE2−0.01
  132. SACS+0.01
  133. TRIM9−0.01
  134. HAPLN1+0.01
  135. C3orf70−0.01
  136. UNC5B−0.01
  137. SLC26A10+0.01
  138. SEMA5B+0.01
  139. PPP1R3C+0.00
  140. MMP16−0.00
  141. PBK+0.00
  142. AFAP1L1−0.00
  143. JPH2−0.00
  144. CHRNA1+0.00
  145. PLA1A+0.00
  146. C6orf223−0.00
  147. C10orf10+0.00
  148. RFX8+0.00
  149. SPC25+0.00
Figure 3. Standardized elastic-net coefficients of the 169 genes selected to separate kidney renal clear cell carcinoma from normal kidney, 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. FKBP101.65 (1.37–1.98)2.9 × 10−6
  2. MICALL21.63 (1.38–1.91)1.9 × 10−7
  3. IQGAP31.54 (1.31–1.82)8.5 × 10−6
  4. IL20RB1.53 (1.34–1.76)1.1 × 10−7
  5. MUC121.51 (1.25–1.83)1.8 × 10−4
  6. ASPM1.51 (1.27–1.79)3.4 × 10−5
  7. SACS1.47 (1.24–1.74)1.1 × 10−4
  8. DIRAS20.78 (0.70–0.87)1.4 × 10−4
  9. SEMA6A0.77 (0.68–0.87)2.0 × 10−4
  10. GDF60.77 (0.68–0.86)1.1 × 10−4
  11. EDA2R0.76 (0.67–0.86)1.7 × 10−4
  12. FATE10.76 (0.67–0.86)1.7 × 10−4
  13. ANO40.76 (0.67–0.86)2.3 × 10−4
  14. MARCH40.75 (0.66–0.85)1.1 × 10−4
  15. RIMKLA0.73 (0.64–0.82)1.5 × 10−5
  16. CXorf360.73 (0.63–0.84)1.1 × 10−4
  17. AFAP1L10.72 (0.63–0.84)1.4 × 10−4
  18. TMEM2330.71 (0.62–0.82)3.4 × 10−5
  19. VWF0.70 (0.61–0.80)1.1 × 10−5
  20. SPATA180.63 (0.56–0.71)1.7 × 10−12
Show the other 39 genes
  1. BMP11.45 (1.21–1.74)3.1 × 10−4
  2. P4HA31.35 (1.16–1.57)6.0 × 10−4
  3. PLOD21.32 (1.12–1.55)0.0037
  4. PCDHGC51.31 (1.10–1.56)0.0088
  5. PADI11.30 (1.12–1.51)0.0032
  6. TBX151.30 (1.10–1.53)0.0080
  7. ARHGEF391.29 (1.09–1.52)0.011
  8. TMEM45A1.28 (1.09–1.51)0.0092
  9. C5orf461.28 (1.08–1.51)0.014
  10. TMEM74B1.28 (1.09–1.50)0.011
  11. FN11.25 (1.07–1.46)0.018
  12. MTCP11.25 (1.07–1.46)0.020
  13. SIX11.22 (1.04–1.43)0.044
  14. NKAIN11.20 (1.03–1.41)0.047
  15. B4GALNT11.19 (1.03–1.37)0.044
  16. PNMA20.86 (0.76–0.96)0.032
  17. SNX330.86 (0.75–0.97)0.049
  18. CDK180.85 (0.75–0.97)0.048
  19. COL23A10.85 (0.75–0.96)0.026
  20. SEMA5B0.85 (0.74–0.96)0.037
  21. HS3ST20.84 (0.73–0.96)0.039
  22. COL4A10.84 (0.73–0.97)0.047
  23. UNC5B0.84 (0.73–0.96)0.039
  24. MYO3A0.83 (0.74–0.94)0.013
  25. FAM163A0.83 (0.72–0.94)0.015
  26. COX4I20.82 (0.71–0.96)0.036
  27. APLN0.82 (0.71–0.94)0.021
  28. NETO20.82 (0.71–0.94)0.015
  29. SLC6A30.81 (0.71–0.93)0.0085
  30. GABRD0.81 (0.70–0.94)0.020
  31. OLFML2A0.81 (0.70–0.93)0.011
  32. SLC28A10.81 (0.72–0.90)8.3 × 10−4
  33. GRIK30.80 (0.70–0.91)0.0026
  34. DLL40.79 (0.69–0.92)0.0079
  35. GPR40.78 (0.68–0.89)0.0021
  36. BARX20.78 (0.69–0.88)6.0 × 10−4
  37. C6orf2230.77 (0.68–0.88)0.0010
  38. PIEZO20.77 (0.67–0.88)0.0014
  39. ATP2B20.74 (0.64–0.85)3.1 × 10−4
longer OSshorter OS
Figure 4. Hazard ratio per standard deviation of tumor expression with 95% confidence interval, log scale, for the 59 genes with FDR < 0.05.

For the 150 top-ranked candidates, a Cox model relates tumor expression, as a continuous value, to overall survival in 528 patients (173 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 overall survival. These are associations in tissue, not evidence that a blood level predicts outcome.

Caveats for KIRC

  • 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