Skip to content
LB2

Liver hepatocellular carcinoma LIHC

Of 60,498 genes, 1,626 are higher in the tumor than in normal liver and in whole blood, and are not made by blood immune cells. 695 are protein-coding, and 693 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
369
TCGA primary tumors
Normal tissue
160
50 TCGA tumor-adjacent, 110 GTEx liver
Whole blood
337
GTEx samples from 328 donors
Survival
364
patients, 179 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

1,626
candidate genes: 693 with plasma cfRNA evidence, 7 protein-coding without, 926 non-coding
21
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. 0.998
60
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 1,626 candidates.

Loading candidates…

How the candidates were selected

60,498 genes tested; 7,055 higher in tumor than in normal liver; 3,399 also higher than in whole blood; 1,626 remain after removing genes expressed in blood immune cells; 693 of these have plasma cfRNA evidence, 7 protein-coding candidates have none, and 926 are non-coding genes the plasma sources cannot list.

Figure 2. Genes remaining after each step in liver hepatocellular 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

Geneweight toward tumorCoef.
  1. SCX+0.58
  2. EBF2+0.51
  3. GABRD+0.44
  4. FATE1+0.32
  5. TIGD5+0.23
  6. COL15A1+0.22
  7. ARHGEF39+0.19
  8. HSF4+0.18
  9. APLN+0.14
  10. E2F1+0.12
  11. MARCH4+0.10
  12. COX4I2+0.09
  13. MARK4+0.08
  14. ZIC2+0.08
  15. PXDNL+0.08
  16. TBCE+0.07
  17. FBXO43+0.06
  18. CD34+0.03
  19. ASPM+0.02
  20. HOXA13+0.01
Show the other 1 genes
  1. MESP1+0.00
Figure 3. Standardized elastic-net coefficients of the 21 genes selected to separate liver hepatocellular carcinoma from normal liver, 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 0.998, 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. NCAPG1.43 (1.22–1.68)0.0011
  2. CDC25C1.41 (1.20–1.65)0.0011
  3. KAT2A1.41 (1.20–1.65)0.0011
  4. GINS11.39 (1.19–1.62)0.0011
  5. FBXO431.38 (1.17–1.63)0.0014
  6. PBK1.37 (1.18–1.60)0.0011
  7. TONSL1.37 (1.17–1.61)0.0011
  8. PLEKHH11.37 (1.16–1.62)0.0015
  9. EXO11.37 (1.17–1.61)0.0014
  10. ATRIP1.37 (1.18–1.59)0.0011
  11. SPC251.35 (1.16–1.58)0.0011
  12. MRGBP1.35 (1.16–1.57)0.0011
  13. FATE11.35 (1.16–1.58)0.0014
  14. ANKRD161.35 (1.16–1.56)0.0011
  15. ASPM1.34 (1.15–1.57)0.0018
  16. KIF141.34 (1.15–1.56)0.0016
  17. XRCC21.34 (1.14–1.56)0.0020
  18. MEX3A1.33 (1.14–1.56)0.0020
  19. DNAH141.33 (1.16–1.53)0.0011
  20. TRIM451.33 (1.15–1.53)0.0011
Show the other 40 genes
  1. ERBB31.44 (1.17–1.77)0.0032
  2. IQGAP31.35 (1.15–1.58)0.0020
  3. ZNF2051.32 (1.11–1.55)0.0055
  4. DSCC11.31 (1.12–1.53)0.0032
  5. DONSON1.31 (1.12–1.52)0.0032
  6. SOGA11.31 (1.12–1.53)0.0042
  7. HSF2BP1.30 (1.11–1.53)0.0058
  8. KIF4A1.30 (1.12–1.51)0.0032
  9. ZNF481.30 (1.12–1.52)0.0038
  10. C16orf591.30 (1.12–1.51)0.0031
  11. CLCN21.29 (1.11–1.50)0.0042
  12. PTDSS21.29 (1.12–1.48)0.0021
  13. HOXD81.29 (1.10–1.51)0.0066
  14. PPP1R371.29 (1.11–1.49)0.0039
  15. FAM86C11.28 (1.09–1.50)0.0079
  16. KIAA15221.28 (1.10–1.48)0.0058
  17. TTL1.27 (1.10–1.48)0.0058
  18. MESP21.27 (1.09–1.49)0.0083
  19. ALMS11.27 (1.09–1.47)0.0079
  20. RPTOR1.27 (1.09–1.47)0.0083
  21. CBX21.26 (1.07–1.47)0.015
  22. RECQL41.26 (1.09–1.45)0.0058
  23. DNMT3B1.25 (1.07–1.46)0.014
  24. ZIC21.25 (1.07–1.48)0.020
  25. ARHGAP391.25 (1.06–1.48)0.022
  26. RCCD11.25 (1.07–1.46)0.014
  27. TEAD31.24 (1.06–1.46)0.020
  28. C19orf481.24 (1.08–1.43)0.0083
  29. PPP1R13L1.23 (1.07–1.42)0.015
  30. ZNF512B1.23 (1.06–1.43)0.023
  31. SOX121.23 (1.05–1.43)0.027
  32. ADCY61.22 (1.05–1.43)0.029
  33. GTF2IRD11.22 (1.04–1.43)0.035
  34. IFT811.22 (1.04–1.42)0.035
  35. LSM111.21 (1.04–1.41)0.035
  36. ZNF6181.21 (1.04–1.41)0.036
  37. SEMA5B1.20 (1.03–1.40)0.043
  38. CYB5RL1.20 (1.03–1.40)0.047
  39. EHMT21.20 (1.04–1.39)0.037
  40. APLN1.20 (1.03–1.39)0.042
longer PFIshorter PFI
Figure 4. Hazard ratio per standard deviation of tumor expression with 95% confidence interval, log scale, for the 60 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 364 patients (179 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 LIHC

  • 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