Skip to content
LB2

Breast invasive carcinoma BRCA

Of 60,498 genes, 590 are higher in the tumor than in normal breast and in whole blood, and are not made by blood immune cells. 375 are protein-coding, and 372 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
1,092
TCGA primary tumors
Normal tissue
292
113 TCGA tumor-adjacent, 179 GTEx breast
Whole blood
337
GTEx samples from 328 donors
Survival
1,077
patients, 151 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

590
candidate genes: 372 with plasma cfRNA evidence, 5 protein-coding without, 213 non-coding
146
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. 0.999
0
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 590 candidates.

Loading candidates…

How the candidates were selected

60,498 genes tested; 1,585 higher in tumor than in normal breast; 1,132 also higher than in whole blood; 590 remain after removing genes expressed in blood immune cells; 372 of these have plasma cfRNA evidence, 5 protein-coding candidates have none, and 213 are non-coding genes the plasma sources cannot list.

Figure 2. Genes remaining after each step in breast invasive 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. COL10A1+0.49
  2. LINC00922+0.45
  3. WISP1+0.44
  4. COL1A2−0.42
  5. ARSI−0.41
  6. ECE2+0.37
  7. BGN+0.36
  8. CHRNA6+0.33
  9. ACAN+0.31
  10. C3orf36+0.27
  11. PPAPDC1A+0.27
  12. COL12A1−0.27
  13. TPRN+0.27
  14. RP11-161H23.5−0.26
  15. KLK4−0.26
  16. ADM2+0.25
  17. RASL11B+0.25
  18. ZACN+0.25
  19. ADAMTS14+0.25
  20. GNG13+0.25
Show the other 126 genes
  1. DNAH14+0.24
  2. SLC24A2+0.24
  3. DRP2−0.23
  4. COMP+0.22
  5. COL3A1−0.21
  6. GABRD+0.20
  7. TMEM54+0.20
  8. C3orf80−0.20
  9. CREB3L1+0.18
  10. GRM4+0.17
  11. ANLN+0.17
  12. ABHD17C+0.17
  13. STRIP2−0.17
  14. WT1+0.17
  15. PPEF1−0.17
  16. FAM72D+0.17
  17. CILP2+0.17
  18. PITX1+0.16
  19. CTXN1+0.16
  20. NPIPA7+0.15
  21. AC093850.2+0.15
  22. DMBX1−0.15
  23. ARHGEF16−0.14
  24. KIF4A+0.14
  25. CASC5+0.14
  26. KCNK15−0.14
  27. HTR1D+0.14
  28. CST2+0.13
  29. COL5A1−0.13
  30. RAB26+0.13
  31. CST1+0.13
  32. LY6H+0.12
  33. INHBA+0.12
  34. COL22A1−0.12
  35. CENPF+0.12
  36. DNMT3B−0.12
  37. MMP13+0.12
  38. ARTN+0.12
  39. GBX2+0.12
  40. TICRR−0.12
  41. COL5A2−0.12
  42. ULBP1+0.11
  43. A1BG+0.11
  44. RECQL4+0.11
  45. ASPM+0.11
  46. ZNF707−0.11
  47. RPLP0P2+0.10
  48. KIF15+0.10
  49. TMEM249+0.10
  50. MAL2−0.10
  51. FAP−0.10
  52. CCL11+0.10
  53. C20orf144+0.09
  54. MATN3+0.09
  55. DQX1+0.09
  56. ZNF367−0.09
  57. EPYC+0.09
  58. KLHL38−0.09
  59. SPC25+0.08
  60. PRR19+0.08
  61. CDH11−0.08
  62. KCNJ10+0.08
  63. PCDH17−0.08
  64. CRABP2−0.08
  65. SLC39A4+0.07
  66. CCDC87−0.07
  67. SPATA17−0.07
  68. HAPLN1+0.06
  69. DSCC1−0.06
  70. GTF2IRD1+0.06
  71. CHRNA1−0.06
  72. RP11-400N13.3+0.06
  73. FOXH1+0.06
  74. CAPSL−0.06
  75. IBSP+0.05
  76. HOXC11+0.05
  77. NEIL3+0.05
  78. LCTL+0.05
  79. SRCIN1−0.05
  80. MMP1+0.04
  81. C19orf48+0.04
  82. FOXO6−0.04
  83. PANO1+0.04
  84. KIAA1211+0.04
  85. SKA1+0.04
  86. CST4+0.04
  87. ACTL10+0.04
  88. CSMD2−0.04
  89. LINGO1−0.04
  90. RNFT2+0.04
  91. MKI67+0.04
  92. RIBC2−0.04
  93. MEX3A−0.03
  94. GSG2−0.03
  95. DNAAF3−0.03
  96. MARCH4−0.03
  97. ENTHD1+0.03
  98. FN1+0.03
  99. KIF14+0.03
  100. HIST2H3A+0.03
  101. HIST2H3C+0.03
  102. LRRC15−0.03
  103. CDCA2+0.03
  104. SYT13−0.02
  105. CADPS−0.02
  106. SCUBE3−0.02
  107. IQGAP3+0.02
  108. MTFR2+0.02
  109. RTN4RL2−0.02
  110. FOXM1+0.02
  111. LHX2+0.02
  112. KLHDC9−0.01
  113. SRMS−0.01
  114. NAT14+0.01
  115. TLE6+0.01
  116. CEMIP+0.01
  117. SPAG1−0.01
  118. RET−0.01
  119. C3orf52−0.01
  120. CAMK2N2+0.01
  121. HSF2BP−0.01
  122. NKX3-2−0.01
  123. VGF+0.00
  124. LRRC73−0.00
  125. SAPCD2+0.00
  126. KCNJ11+0.00
Figure 3. Standardized elastic-net coefficients of the 146 genes selected to separate breast invasive carcinoma from normal breast, 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 0.999, 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

None of the 150 top-ranked candidates is associated with overall survival at FDR < 0.05 (1,077 patients, 151 events). A null result here is informative: these genes mark the presence of the cancer, not necessarily its course.

For the 150 top-ranked candidates, a Cox model relates tumor expression, as a continuous value, to overall survival in 1,077 patients (151 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 BRCA

  • 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