Skip to main content
AACR Project GENIE v19.0 · 21,017 myeloid samples Panel-adjusted Fisher's exact with Benjamini-Hochberg FDR N=1 case study · Not clinical guidance

Co-occurrence Analysis

190 gene-pair Fisher's exact tests, Benjamini-Hochberg FDR, quintuple search

Gene Pairs Tested
190
Significant (BH)
145
p<0.05 after correction
Mutual Exclusions
30
Strongest
FLT3+NPM1
O/E = 9.24
Quintuple Frequency
< 1 in 2,889
Exact 95% bound from 0 of 8,655 patients; no point estimate

Co-occurrence Heatmap

Source: Fisher's exact test 31
[31] C 2013
Mutational landscape and significance across 12 major cancer types. Nature (2013)
, Benjamini-Hochberg FDR correction. GENIE v19.0 17
[17] Consortium 2017
AACR Project GENIE: Powering Precision Medicine through an International Consortium. Cancer Discov (2017)
, 21,017 myeloid samples (sample level; selected by cancer-type keywords with hypermutation cut-off 40, an older filter than the 21,017-sample OncoTree cohort used elsewhere). Exclusion means O/E below 1 in the same two-sided Fisher's exact test; the DISCOVER test 30
[30] S 2016
A novel independence test for somatic alterations in cancer shows that biology drives mutual exclusivity but chance explains most co-occurrence. Genome Biol (2016)
, which allows mutation rates to vary between samples, was not applied here. Color scale: log2(O/E). Blue = mutual exclusion, red = co-occurrence.

Five-gene Co-mutation Landscape (GENIE v19.0 OncoPrint)

Figure: GENIE v19.0 myeloid patients with all five genes on the sequencing panel (8,655; samples merged per patient; myeloid OncoTree, coding variants only, hypermutation filter). Each column is one patient with coding mutations in at least 3 of DNMT3A, IDH2, SETBP1, PTPN11, EZH2 (62 patients: 60 with three genes, 2 with four, 0 with all five). Gene level, not the exact variants. Right: percentage of the 8,655 patients with a coding mutation in the gene. Source: oncoprint_matrix.json (syn72382128).

Top Co-occurrences (sample level)

Gene PairObserved (samples)Expected (samples)O/Ep-valueBH q-valueDirection
FLT3 + NPM142445.99.24< 1e-300< 1e-300Co-occurring
ASXL1 + SRSF2737228.43.237.51e-2297.13e-227Co-occurring
TET2 + SRSF2755248.33.042.42e-2171.53e-215Co-occurring
SRSF2 + RUNX1561153.23.666.00e-1942.85e-192Co-occurring
ASXL1 + RUNX1705236.82.981.51e-1915.74e-190Co-occurring
ASXL1 + EZH239693.94.222.84e-1678.98e-166Co-occurring
DNMT3A + NPM1443117.63.774.01e-1631.09e-161Co-occurring
SRSF2 + IDH237986.84.371.49e-1553.53e-154Co-occurring
ASXL1 + SETBP128957.05.077.92e-1521.67e-150Co-occurring
ASXL1 + STAG234479.14.353.80e-1517.21e-150Co-occurring
SRSF2 + STAG225948.95.291.03e-1261.78e-125Co-occurring
RUNX1 + EZH227763.04.401.51e-1112.39e-110Co-occurring
TET2 + ASXL1768387.71.988.07e-951.18e-93Co-occurring
DNMT3A + FLT3396139.92.832.79e-943.79e-93Co-occurring
RUNX1 + BCOR24057.54.172.54e-903.21e-89Co-occurring

SETBP1 Partnerships (patient level)

PartnerCo-mutated (patients)Expected (patients)O/Ep-valueBH q-valueDirection
ASXL114531.14.674.40e-711.45e-69Co-occurring
SRSF27921.13.736.37e-271.05e-25Co-occurring
CBL326.64.857.09e-147.80e-13Co-occurring
CSF3R181.810.001.10e-139.08e-13Co-occurring
EZH2368.84.073.80e-132.51e-12Co-occurring
U2AF13911.33.465.75e-123.16e-11Co-occurring
RUNX13920.01.954.18e-51.97e-4Co-occurring
NRAS2913.32.196.92e-52.86e-4Co-occurring
JAK21433.20.428.32e-53.05e-4Exclusive
NPM1 Depleted212.60.163.55e-40.0012Exclusive
PTPN11156.02.519.60e-40.0029Co-occurring
GATA2114.32.540.00410.0113Co-occurring
IDH1 Depleted17.80.130.00510.0130Exclusive
TP531425.40.550.01360.0322Exclusive
SMC341.23.250.03390.0734Co-occurring
BCORL104.00.000.03560.0734Exclusive
DDX4120.54.260.07860.1526Co-occurring
STAG2116.61.680.10190.1867Co-occurring
SF3B1813.60.590.12000.2040Exclusive
WT137.10.420.12360.2040Exclusive
CALR47.60.530.25780.4051Exclusive
RAD2142.31.730.29460.4418Co-occurring
ZRSR264.11.450.31090.4461Co-occurring
CEBPA24.50.440.33180.4562Exclusive
FLT31012.40.810.55620.7341Exclusive
SMC1A01.30.000.63860.8105Exclusive
TET23436.50.930.71430.8469Exclusive
BCOR98.21.090.71850.8469Co-occurring
DNMT3A3028.91.040.83980.9364Co-occurring
KRAS87.61.050.85120.9364Co-occurring
IDH21213.10.920.88530.9424Exclusive
MPL33.60.841.00001.0000Exclusive
PHF655.40.931.00001.0000Exclusive
IDH1 depletion (O/E=0.13) and NPM1 depletion (O/E=0.17) stand out as the strongest mutual exclusions with SETBP1 mutations 1
[1] R 2013
Recurrent SETBP1 mutations in atypical chronic myeloid leukemia. Nat Genet (2013)
2
[2] H 2013
Somatic SETBP1 mutations in myeloid malignancies. Nat Genet (2013)
. This is biologically consistent: SETBP1-mutant clones occupy a distinct ontogenetic niche (MDS/MPN overlap, CNL, aCML) where NPM1 mutations are rare and IDH1 mutations preferentially co-occur with NPM1 in AML 12
[12] E 2016
Genomic Classification and Prognosis in Acute Myeloid Leukemia. N Engl J Med (2016)
. The strongest positive associations are CSF3R (O/E=10.00, p=1.1 × 10−13) and ASXL1 (O/E=4.67, p=4.4 × 10−71), reflecting the classical CNL/aCML genotype.
The IDH1+SETBP1 depletion (O/E=0.13) is consistent with SETBP1's role as an epigenetic hub 3
[3] R 2018
SETBP1 induces transcription of a network of development genes by acting as an epigenetic hub. Nat Commun (2018)
, where mutant SETBP1 drives transcriptional deregulation via HCF1/KMT2A interaction and PP2A inhibition. Redundant epigenetic disruption through IDH1-mediated DNA hypermethylation appears to be negatively selected when SETBP1 already provides broad chromatin remodeling. This functional overlap explains the mutual exclusivity: clones carrying both mutations gain no additional fitness advantage, so the combination is purged by clonal competition.

Quintuple Co-occurrence Analysis

The quadruple analysis assumed four driver mutations. With EZH2 V662A classified Likely Pathogenic (PP3_Strong on 6/7 concordant computational predictors; EVE dissents at 0.7825, Uncertain at its class25 calibration), this becomes a quintuple co-occurrence search. The exact five-variant combination has no estimable population frequency. Zero in 8,655 patients bounds it below 1 in 2,889 (95%, exact binomial).
EZH2 V662A has 0 carriers among the 20,739 GENIE v19.0 17
[17] Consortium 2017
AACR Project GENIE: Powering Precision Medicine through an International Consortium. Cancer Discov (2017)
myeloid samples whose panel covers EZH2. The variant itself does not exist in the database, making any multi-gene search involving this specific variant a guaranteed zero. Because that count is zero, the frequency of all five specific variants (DNMT3A R882H + IDH2 R140Q + SETBP1 G870S 1
[1] R 2013
Recurrent SETBP1 mutations in atypical chronic myeloid leukemia. Nat Genet (2013)
2
[2] H 2013
Somatic SETBP1 mutations in myeloid malignancies. Nat Genet (2013)
+ PTPN11 E76Q + EZH2 V662A) in a single myeloid patient cannot be estimated under any model: every model value depends on the number put in for that zero. Zero in 8,655 patients bounds it below 1 in 2,889 (95%, exact binomial). A pairwise maximum-entropy (Ising) model fitted to GENIE predicts mutations in all five genes in about 1 in 25,000 myeloid patients (95% bootstrap interval 1 in 13,000 to 1 in 65,000; sampling error only).
CombinationObservedExpectedNotes
DNMT3A p.R882H + IDH2 p.R140Q + SETBP1 p.G870S00.21No triple match in GENIE (expected count in 18,628 eligible samples)
DNMT3A p.R882H + IDH2 p.R140Q + PTPN11 p.E76Q00.02No triple match in GENIE (expected count in 18,644 eligible samples)
DNMT3A p.R882H + SETBP1 p.G870S + PTPN11 p.E76Q02.5e-3No triple match in GENIE (expected count in 19,974 eligible samples)
IDH2 p.R140Q + SETBP1 p.G870S + PTPN11 p.E76Q02.8e-3No triple match in GENIE (expected count in 18,625 eligible samples)
All four variants (original)01.1e-4Expected count in 18,625 eligible samples
EZH2 V662A (fifth driver)000/20,739 panel-eligible samples; variant absent from GENIE
Gene-level quintuple (any variant)00.0034Any DNMT3A+IDH2+SETBP1+PTPN11+EZH2 variant; expected count under independence. A fitted Ising model gives about 1 in 25,000 patients (0.34 expected in 8,655)
All five specific variants0Not estimable95% bound below 1 in 2,889 (0 of 8,655 patients, exact binomial)
Source: Expected values are simple products of frequencies under independence, without the DISCOVER correction for varying mutation rates 30
[30] S 2016
A novel independence test for somatic alterations in cancer shows that biology drives mutual exclusivity but chance explains most co-occurrence. Genome Biol (2016)
: P(quint) = P(DNMT3A R882H) x P(IDH2 R140Q) x P(SETBP1 G870S) x P(PTPN11 E76Q) x P(EZH2 V662A). Individual variant frequencies from GENIE v19.0 17
[17] Consortium 2017
AACR Project GENIE: Powering Precision Medicine through an International Consortium. Cancer Discov (2017)
myeloid panel-eligible cohort (N=21,017 myeloid, N=18,625 panel-eligible for original four). EZH2 V662A frequency = 0/20,739 panel-eligible samples. Gene-level quintuple uses gene-level (any coding variant) frequencies. The five-variant row shows the exact 95% bound instead of an expected value, because EZH2 V662A has 0 carriers; bound and Ising model from rarity-reanalysis.json.
References
  1. Piazza R et al. Recurrent SETBP1 mutations in atypical chronic myeloid leukemia. Nat Genet (2013). PubMed
  2. Makishima H et al. Somatic SETBP1 mutations in myeloid malignancies. Nat Genet (2013). PubMed
  3. AACR Project GENIE Consortium. AACR Project GENIE: powering precision medicine through an international consortium. Cancer Discov (2017). DOI
  4. Kandoth C et al. Mutational landscape and significance across 12 major cancer types. Nature (2013). PubMed
  5. Canisius S et al. A novel independence test for somatic alterations in cancer. Genome Biol (2016). DOI
  6. Papaemmanuil E et al. Genomic classification and prognosis in acute myeloid leukemia. N Engl J Med (2016). PubMed
  7. Piazza R et al. SETBP1 induces transcription of a network of development genes by acting as an epigenetic hub. Nat Commun (2018). PubMed