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

Clonal Architecture

In-house reimplementation of the PyClone-VI model, CCF estimates and tree comparison

Preferred Model Favored (heuristic score)
Linear
Score -0.51 vs -14.11 (branching); not a formal likelihood i
Founder Clone CCF 1.00 / 0.92
DNMT3A / EZH2
Order unresolved: 95% CIs overlap (0.88 to 1.00 vs 0.85 to 0.98); DNMT3A CCF 1.00 is set by the purity estimate i
Late Subclone CCF 0.05
IDH2
R140Q, 2% VAF, only druggable target (enasidenib) i
Tumor Purity
0.78
Cross-validated: DNMT3A 0.78, EZH2 (monosomy 7) 0.74 i
REVOLVER Cohort
7,956 samples
4,971 patients. Input matrix prepared; REVOLVER not yet run

Cancer Cell Fraction Estimates

Source: In-house reimplementation of the PyClone-VI model (beta-binomial CCF posterior, Dirichlet Process Mixture clustering; each mutation formed its own cluster) 16
[16] S 2020
PyClone-VI: scalable inference of clonal population structures using whole genome data. BMC Bioinformatics (2020)
. Purity estimated from highest diploid VAF (DNMT3A R882H at 39%).

CCF Detail

GeneVariantVAFCCF95% CIPathway
EZH2V662A0.590.92280.854 to 0.984Polycomb complex (epigenetic)
DNMT3AR882H0.391.00000.878 to 0.999DNA methyltransferase (epigenetic regulation)
SETBP1G870S0.340.87180.769 to 0.973MDS/MPN overlap, PP2A inhibition
PTPN11E76Q0.290.74360.646 to 0.849RAS-MAPK signaling (gain-of-function)
IDH2R140Q0.020.05130.028 to 0.093Metabolic (2-HG production)
CCF computed as VAF * (purity * total_cn + normal_cn * (1-purity)) / (purity * mutant_cn). EZH2 on monosomy 7 (hemizygous).

Linear vs Branching Model

ModelHeuristic scoreInterpretation
Linear (sequential)-0.51Favored by a heuristic score (clones sorted by CCF, no formal likelihood). Top four clones must be nested; IDH2 placement open.
Branching (parallel)-14.11Lower score (fixed penalties for branching). Not excluded for the IDH2 subclone.
The linear model scores higher (heuristic score -0.51) than branching (-14.11). The score is not a formal likelihood. Ordered by CCF, the clones are:

DNMT3A R882H (CCF 1.00) → EZH2 V662A (0.92) → SETBP1 G870S (0.87) → PTPN11 E76Q (0.74) → IDH2 R140Q (0.05)

The top four clones (CCF 0.74 to 1.00) must be nested, since any two of their CCFs sum to more than 1. DNMT3A and EZH2 cannot be ordered (overlapping intervals), so the founding clone carries DNMT3A R882H and/or EZH2 V662A on monosomy 7, both near-clonal, before SETBP1 G870S. IDH2 (CCF 0.05) may sit inside the PTPN11 clone or on a separate branch.

Type 1 MDS-to-AML Progression

The clonal trajectory matches the Type 1 progression pattern defined by Makishima et al. 15
[15] H 2017
Dynamics of clonal evolution in myelodysplastic syndromes. Nat Genet (2017)
: Type 1 mutations, here PTPN11 (signaling) and IDH2 (metabolic), are acquired during MDS-to-AML transformation, layered on top of the founding epigenetic lesions.

PTPN11 E76Q (CCF 0.74) provides the proliferative RAS-MAPK signal for blast expansion. IDH2 R140Q (CCF 0.05) represents a late subclonal event producing 2-hydroxyglutarate, the only FDA-approved druggable target in this profile (enasidenib 21
[21] EM 2017
Enasidenib in mutant IDH2 relapsed or refractory acute myeloid leukemia. Blood (2017)
).

The therapeutic paradox: the druggable mutation (IDH2) is the smallest subclone. Eliminating it with enasidenib would treat ~5% of the tumor mass. The dominant clones carrying DNMT3A, EZH2, SETBP1, and PTPN11 have no direct targeted therapy.
PhaseMutationCCFRole
Founding / early (order with EZH2 unresolved)DNMT3A R882H1.00Epigenetic initiator (dominant-negative methyltransferase)
Founding / early (order with DNMT3A unresolved)EZH2 V662A0.92Chromatin remodeling loss-of-function (PRC2 axis)
MDS bridgeSETBP1 G870S0.87PP2A inhibition, bridges epigenetic and proliferative axes
TransformationPTPN11 E76Q0.74RAS-MAPK gain-of-function, blast proliferation signal
Late subcloneIDH2 R140Q0.05Metabolic (2-HG), only druggable target
Type 1 progression pattern per Makishima et al. Nat Genet 2017.

Triple Epigenetic Catastrophe

The patient carries three convergent epigenetic regulators. Two (DNMT3A, EZH2) are near-clonal; the third (IDH2) is a late minor subclone:

1. DNMT3A R882H (CCF 1.00): dominant-negative loss of DNA methyltransferase activity: the R882H enzyme has about 80% less methyltransferase activity than wild type, with focal CpG hypomethylation in AML cells (Russler-Germain et al., Cancer Cell 2014).

2. EZH2 V662A (CCF 0.92): predicted loss-of-function (in-silico scores; V662A itself not tested) in the PRC2 catalytic SET domain, where missense mutations abolish H3K27 methylation in earlier in vitro assays cited by Chase et al. 10
[10] A 2020
Mutational mechanisms of EZH2 inactivation in myeloid neoplasms. Leukemia (2020)
.

3. IDH2 R140Q (CCF 0.05): neomorphic production of 2-hydroxyglutarate (2-HG), which inhibits TET2 and other alpha-ketoglutarate-dependent dioxygenases. Though subclonal, even low levels of 2-HG can disrupt the epigenetic landscape globally.

Together, these mutations attack three independent arms of the epigenetic machinery: DNA methylation (DNMT3A), histone methylation (EZH2/PRC2), and metabolic regulation of both (IDH2/TET2). No existing treatment protocol is designed to address this convergent epigenetic collapse. The resulting chromatin state is likely profoundly deregulated, creating a permissive environment for the signaling mutations (SETBP1, PTPN11) that drive transformation.

REVOLVER Evolutionary Trajectory Cohort

A binary mutation matrix of 7,956 myeloid samples (4,971 patients; samples with at least one coding mutation in the 34 target genes) has been prepared for REVOLVER (Repeated Evolution in Cancer) analysis. This matrix enables inference of repeated evolutionary trajectories across the GENIE v19.0 myeloid cohort.

The patient's linear trajectory (DNMT3A → EZH2 → SETBP1 → PTPN11 → IDH2) can be compared against population-level evolutionary patterns to determine whether this ordering is recurrent or unique. REVOLVER has not been run yet, so no population ordering is available. The patient-level ordering comes from the in-house reimplementation of the PyClone-VI model 16
[16] S 2020
PyClone-VI: scalable inference of clonal population structures using whole genome data. BMC Bioinformatics (2020)
, where the linear tree scores higher on a heuristic score.
ParameterValue
Total samples7,956 (4,971 patients)
Genes34 myeloid driver genes
Matrix formatBinary (mutated/wildtype)
Cohort sourceGENIE v19.0, myeloid-filtered
IDH2-SETBP1 exclusivity (GENIE Fisher test, not a REVOLVER result)Not significant (O/E 0.905, p = 0.74)
REVOLVER input matrix prepared from GENIE v19.0 myeloid cohort. REVOLVER itself has not been run.
References
  1. Gillis S, Roth A. PyClone-VI: scalable inference of clonal population structures using whole genome data. BMC Bioinformatics (2020). DOI
  2. Makishima H, Yoshizato T, Yoshida K, et al. Dynamics of clonal evolution in myelodysplastic syndromes. Nat Genet (2017). PubMed
  3. Chase A, Score J, Lin F, et al. Mutational mechanisms of EZH2 inactivation in myeloid neoplasms. Leukemia (2020). PubMed
  4. Stein EM, DiNardo CD, Pollyea DA, et al. Enasidenib in mutant IDH2 relapsed or refractory acute myeloid leukemia. Blood (2017). PubMed