Interplay between chromosomal alterations and gene mutations shapes the evolutionary trajectory of c

Stably acquired mutations in hematopoietic cells represent substrates of selection that may lead to clonal hematopoiesis (CH), a common state in cancer patients that is associated with a heightened risk of leukemia development. Owing to technical and sample size limitations, most CH studies have characterized gene mutations or mosaic chromosomal alterations (mCAs) individually. Here we leverage peripheral blood sequencing data from 32,442 cancer patients to jointly characterize gene mutations (n = 14,789) and mCAs (n = 383) in CH. Recurrent composite genotypes resembling known genetic interactions in leukemia genomes underlie 23% of all detected autosomal alterations, indicating that these selection mechanisms are operative early in clonal evolution. CH with composite genotypes defines a patient group at high risk of leukemia progression (3-year cumulative incidence 14.6%, CI: 7–22%). Multivariable analysis identifies mCA as an independent risk factor for leukemia development (HR = 14, 95% CI: 6–33, P < 0.001). Our results suggest that mCA should be considered in conjunction with gene mutations in the surveillance of patients at risk of hematologic neoplasms.

Genome instability is a hallmark of aging and cancer1. Recent studies show that stably acquired somatic mutations occur across human tissues throughout life, in an age-dependent manner. A subset of these mutations drive the establishment of “benign” macroscopic clones2,3,4 and some may lead to overt neoplastic disease. This phenomenon has been best studied in the context of clonal hematopoiesis (CH)5,6 due to the relative ease of blood sampling from healthy individuals. Large-scale population-based studies have established a detailed understanding of CH prevalence and mutation characteristics, as well as the incidence of clinical sequelae such as hematologic neoplasms7,8,9,10,11,12.

Most CH studies focused on gene mutations including single nucleotide substitutions and indels. CH defined by mosaic chromosomal alterations (mCAs), such as amplifications, deletions, and copy-neutral loss of heterozygosity (CNLOH), has also been reported, albeit at a much lower prevalence13,14,15,16,17,18. In both definitions, CH exhibits a strong age-related incidence and shares similar mutational characteristics with hematological cancers. While gene mutations that target established genes in leukemia pathogenesis are postulated to confer an increased fitness of hematopoietic stem and progenitor cells, the mechanisms of selection driving most mCA events are less well understood. Recent studies show that a proportion of mCAs gain selective advantage through interactions with germline alleles13,17,18 and allude to a similar mechanism for pre-existing somatic mutations. However, the relationship between acquired gene mutations and mCAs in CH has not been systematically investigated. Progress in this regard has been limited by the lack of statistically powered studies that simultaneously map chromosomal alterations and gene mutations in the same sample7,9,13,14,15,16,17,19 (Supplementary Table 1).

CH is linked to an increased risk of subsequent hematologic disease suggesting that it likely represents an early step in leukemic transformation. However, most CH cases do not progress to overt disease. Putative driver (PD) status, mutation burden and clone size as estimated by variant allele fraction (VAF) of acquired gene mutations are established predictors of disease progression in CH10,11. Such criteria are increasingly being considered clinically to evaluate individuals at risk of leukemic transformation20,21,22,23. Small-scale retrospective studies have demonstrated that mCAs can either be simultaneously detected with gene mutations or represent the sole detectable abnormalities years before leukemia development10,24. However, the clinical significance of mCAs in the context of co-occurring gene mutations has not been investigated.

Recent studies showed that solid tumor cancer patients have high rates of CH and are at an increased risk of secondary leukemias25,26,27. Here we seek to understand the relationship between mCAs, acquired gene mutations, and the incidence of hematological neoplasms in patients with cancer. We develop a method to reliably detect mCAs in data from established prospective sequencing assays used in routine clinical practice. We apply this method in a cohort of 32,442 cancer patients and characterize in detail the incidence and presentation of mCAs in the context of concurrent gene mutations. We provide evidence of direct positive selection mediated by synergistic effects between gene mutations and chromosomal alterations. We show that mCAs are independent predictors of progression to hematological neoplasms and demonstrate the value of incorporating mCAs in the reporting and interpretation of CH in cancer patients22,23.

Patient cohort and samples

The study population included 32,442 solid tumor patients (median age: 61, range 20–99, Supplementary Table 2) with non-hematologic cancers that underwent matched tumor and blood sequencing at MSKCC using the MSK-IMPACT panel28 on an institutional prospective sequencing protocol (ClinicalTrials.gov number, NCT01775072) before February 1, 2020 (Supplementary Fig. 1). The majority of our samples were sequenced by the most recent version of the panel, IMPACT6, which captures the exonic regions of 468 recurrent cancer genes and includes probes against 1042 common single nucleotide polymorphisms (SNPs). This effectively mimics a low-density SNP tiling array with markers evenly distributed across the genome. Clinical annotations including age, gender, race, primary cancer diagnosis, length of follow-up, serial blood counts, and smoking history were available for most patients in the study. For 10,375 patients complete treatment data was previously collected26. The patients in this study had no record of hematologic malignancies within 3 months following sequencing (Methods).

Landscape of mCAs in prospective sequencing of cancer patients

To study the prevalence of mCAs in patients who have undergone MSK-IMPACT testing, we developed a method, FACETS-CH (a modification of FACETS29) that is optimized for detecting chromosomal alterations at low cell fractions from deep targeted sequencing data. The precision of this approach was validated in 919 technical replicates (Methods, Supplementary Figs. 2 and 3). We applied this method to map chromosomal alterations in the matched blood of 32,442 solid tumor patients in our study. We detected 383 mCA events (335 autosomal) in 346 individuals. The incidence of mCA increases with age, from under 1% among individuals under 50 to more than 3% among subjects who were above 80 (Fig. 1a). We used the magnitude of deviations in logR and logOR (Methods) to estimate the aberrant cell fraction and classify events based on copy number states (Fig. 1b); 156 (47%) autosomal events were classified as deletions, 123 (37%) as CNLOH, and 56 (17%) as gains. Among mCA individuals, 317 (92%) harbored one event, whereas 29 (8%) had multiple (range 1–4) events (Fig. 1c). The smallest clone that we detected had an estimated cell fraction of 10% (Fig. 1d), a limit of detection comparable to prior studies based on high-density SNP-arrays without haplotype phasing (Supplementary Fig. 4)14,15,16. The genomic distribution and patterns of the mCAs we detected using a targeted sequencing platform with sparse genome-wide coverage were broadly consistent with findings from high-density SNP array studies, confirming the specificity of our approach13,14,15,16 (Fig. 1e and Supplementary Figs. 5–8). Overall, patients with mCA showed a similar blood count distribution and cytopenia incidence as ones without mCA (Supplementary Fig. 9).

Among different cancer types in our study, mCA was prevalent in soft tissue sarcoma, thyroid, and lung cancer patients and underrepresented in prostate and bladder cancer patients (Fig. 1f). In a multivariable regression model, mCA was significantly associated with age (OR = 1.8, P < 0.001), male gender (OR = 1.3, P = 0.012), and white race (OR = 1.5, P = 0.033), but not smoking (OR = 1.2, P = 0.18), which was consistent with findings from prior studies16 (Supplementary Fig. 10). For a subset of 10,375 patients whose complete clinical treatment history had been previously collected26, we explored the association of mCA with receipt of oncologic therapy. We found that mCA was positively associated with external beam radiation therapy (OR = 1.7, P = 0.022) but not cytotoxic chemotherapy (OR = 0.9, P = 0.56) (Supplementary Fig. 10). It is likely that differences in exposures to oncologic therapies may partially account for the distinct mCA representation in different tumor types. However, the association between therapies and mCAs is likely to be heterogeneous26. Larger datasets with more complete treatment annotation are needed to fully assess the association between specific subclasses of therapy and specific types of mCAs.

Global characteristics of mCAs in relation to gene mutations

To evaluate the relationship between mCAs and acquired gene mutations, we characterized gene mutations in our patient cohort using an established variant calling procedure from our previous studies25,26. A total of 14,789 mutations were detected in 9854 (30%) individuals across 457 genes captured in our clinical sequencing assay28 (Supplementary Fig. 11a, b). The incidence of gene mutations increased from 5% among patients in their 20s to more than 60% in elderly patients aged above 80 (Supplementary Fig. 12b). Among the mutation-positive patients, 6919 (70%) patients harbored a single mutation while 2935 (30%) harbored 2 or more (Supplementary Fig. 11c). The estimated clone sizes (median 9%) of these mutations were lower than mCAs detected in this study, which reflects our higher detection sensitivity (VAF ≥ 2%) for gene mutations as compared to mCAs at low cell fractions (Supplementary Fig. 12a). The estimates of mutation prevalence are dependent on the detection sensitivity and genomic coverage specific to the sequencing assay used in this study as well as characteristics of the MSK-IMPACT patient cohort.

Considering both variant classes, 217 (63%) CH cases with mCAs co-occurred with at least one gene mutation captured by our panel, while 129 (37%) did not (Fig. 2a; Fisher’s exact test, OR = 3.9, P < 0.001). mCA was especially enriched in CH cases with high mutation number and VAF (Fig. 2b, c). The age distributions for the two types of CH when found in isolation were comparable while the acquisition of both shifted towards older age (Fig. 2d). However, the high degree of overlap between subjects with mCA and gene mutation could not be simply explained by a shared age-related incidence (P < 0.001, regression adjusted for age). Taken together, the high degree of co-occurrence of chromosomal aberrations and gene mutations points towards a potential synergistic relationship. This would be consistent with a multi-hit process of carcinogenesis, where the first hit provides a fertile ground for the acquisition of subsequent hits that leads to further selection and expansion of mutant clones.

Patterns of co-occurrence between mCAs and gene mutations reveal diverse mechanisms of selection

Population-based studies of leukemia genomes have delivered a detailed understanding of recurrent genetic interactions in overt neoplastic disease30,31. Such studies show that chromosomal alterations can be found in cis of existing mutations (i.e., overlapping the mutated locus), thereby contributing to an increase in mutant allele dosage, e.g., 9pCNLOH and JAK2 in myeloproliferative neoplasms (MPN) or bi-allelic inactivation of tumor suppressors, e.g., TP53 and 17pLOH32. Additionally, significant co-occurrences have been reported between specific chromosomal alterations and gene mutations in trans (i.e., not overlapping the mutated locus) which suggest functional co-operativity31,33,34,35,36,37. Furthermore, the emergence of chromosomal alterations can result from global genomic instability induced by loss of gatekeeper function such as genes (e.g., TP53) in the DNA damage response (DDR) pathway31,32,36. The enrichment of composite genotypes observed in this study suggests that these genetic interactions may already be in play at the initial stage of somatic evolution in the absence of overt disease.

CH subjects with mCA had a distinct spectrum of gene mutations as compared to the remaining cohort (Fig. 2e, f). Notably, a large proportion of the genes enriched in patients with mCA (DNMT3A, TET2, TP53, ATM, JAK2, EZH2, RUNX1, MPL) are also frequently affected by loss of heterozygosity in hematopoietic neoplasms32,38. We therefore asked whether the presence of gene mutations is associated with chromosomal alterations in cis. Overall, 2% non-coding variants that co-occurred with an mCA event localized in cis with the mCA (permutation P = 0.78), whereas it is 15% for the coding variants (permutation P < 0.001; Supplementary Fig. 13). This enrichment can be accounted for by mutations in seven genes that recurrently co-localized with an mCA (Fig. 3a, b). Of six events of CNLOH on chr7q, all six co-localized with an EZH2 (7q36.1) mutation (q < 0.001). The high VAFs and the absence of 7qCNLOH events without a concurrent EZH2 mutation suggest that 7qCNLOH is a highly directed event targeting a previously acquired mutation of EZH2 in CH (Fig. 3b and Supplementary Fig. 14a). Of 12 cases with 9pCNLOH, 11 (92%, q < 0.001) co-localized with a JAK2 V617F mutation, while 4 out of 9 (44%, q < 0.001) 1pCNLOH events co-localized with a MPL (1p34.2) mutation. These events mirror cis dosage adjustment effects identified in studies of MPN, here identified in patients without active disease. Additionally, we identified recurrent bi-allelic targeting of TET2 mutations by CNLOH events as well as focal deletions (Fig. 3b). The same is true of chromosomal alterations spanning DNMT3A (deletions), ATM, and TP53 (deletions and CNLOH) (Fig. 3b). The cellular fractions of these cis multi-hit events generally supported homozygosity of the mutant allele and shared clonal origin (Supplementary Fig. 14a).

We next evaluated co-mutations of mCA and gene mutation in trans, which could reflect synergistic effects in cell proliferation or loss of gateway effects promoting global genomic instability31,32,36. We scanned for pairwise associations of specific chromosomal alterations with genes recurrently mutated in patients with mCA (Fig. 3c). These analyses revealed highly specific co-mutational patterns reflective of known genetic interactions in a wide range of blood malignancies30,31,32,37,39. SRSF2/TET2 are commonly co-mutated in myelodysplastic syndromes (MDS) characterized by myelodysplasia and monocytosis40. In our data, 2 out of 7 (29%) CNLOH and 3 out of 14 (21%) deletions targeting 4q24 co-occurred with mutations in SRSF2. Similarly, 3 out of 6 (50%) CNLOH on chr7q co-occurred with a mutation in ASXL1, while deletions on chr7 also recurrently co-mutated with PPM1D. Mutations in U2AF1, albeit rare, co-occurred with 3 out of 44 (7%) cases of 20q deletions (Fig. 3c). These combinations have been previously reported as significantly recurrent in myeloid disease suggesting that they are intrinsically implicated in myeloid pathogenesis33,41. Similarly, we also observe co-mutations characteristic of chronic lymphocytic leukemia (CLL); 6 out of 16 (37.5%) cases of trisomy 12 co-occurred with a mutation in NOTCH1, MYD88, or FBXW7; 2 out of 9 (22%) deletions on chr13q co-occurred with a ATM mutation34. Last, a key part of the DNA damage response pathway, TP53 acts as a gatekeeper for genomic instability in a wide range of cancers. We found that CH mutations in TP53 were associated with numerous chromosomal alterations (deletions in chr5, chr7, and trisomy chr3) (Fig. 3c). In addition, we observe a significant association of TP53 mutations with the presence of multiple concomitant mCAs in CH patients (Fisher’s exact test, P = 0.0015, Supplementary Fig. 15).

Taken together, acquired gene mutations in cis formed putative “double hits” with 26% of all CNLOH events and 8% of deletions in the study. Recurrent co-mutations in trans and loss of gatekeeper function (i.e., TP53) underlied an additional 13% of deletions and 20% of amplifications. In total, 23% of all autosomal mCA events mapped in this study are subject to these recurrent co-mutations. Interestingly, most of the subjects with recurrent CH co-mutations did not exhibit significant blood abnormalities at the time of blood draw (Supplementary Fig. 16). It is also of note that certain mCAs (e.g., chr20-, X-) frequently occurred in isolation without any clear co-mutations, indicating that these events can also drive clonal expansions alone (Fig. 3a). While the mechanisms driving CH are becoming increasingly understood for each respective class of mutations, here by integrating chromosomal alterations with gene mutations we unravel previously under-characterized patterns of mutation acquisition reflective of diverse mechanisms of selection operating in CH.

Evolution to hematologic malignancies

The discovery of composite genotypes characteristic of hematologic neoplasms in patients without an active disease raises the possibility that they represent markers of incipient leukemic transformation. We reviewed hospital billing records and tumor registry data and identified 96 patients with a confirmed hematological neoplasm diagnosis dated more than 3 months after blood collection (Methods, Supplementary Fig. 17). Of these patients (median time to diagnosis 17 months, range 3–62 months) 26 were diagnosed with MDS, 11 with acute myeloid leukemia (AML), 7 with MPN, 4 with chronic myeloid leukemia (CML), 12 with CLL, 27 with B-cell non-Hodgkin’s lymphoma (B-NHL), 2 with T-cell non-Hodgkin’s lymphoma (T-NHL), and 7 with multiple myeloma (MM) (Fig. 4a and Supplementary Fig. 18).

Among the 96 patients who received a diagnosis, 40 (42%) had evidence of only gene mutation at the time of blood profiling, 17 (18%) had both gene mutation and mCA, whereas 1 (1%) had mCA only (Fig. 4a). Overall, the mutation patterns observed at blood draw were characteristic of the respective disease types for patients who subsequently developed MDS and MPN, but less so for the remaining disease subsets that are frequently characterized by genomic rearrangements (MM, B-NHL, T-NHL). Among the patients who subsequently developed MDS, those with mCA appeared to have a shorter time to diagnosis (median time to transformation 9.7 months, range: 4.2–38.7) as compared to patients without mCA (median time 23.2 months, range 9–61.8). Likewise, of the 7 patients who developed MPN, the first 5 to be diagnosed (within 22 months) had concomitant JAK2/9pCNLOH or MPL/1pCNLOH events. This is consistent with studies linking JAK2 homozygosity to increased disease severity in MPN42. In some cases of lymphoma and MM, gene mutations in DNMT3A or TET2 were the only detected events (Fig. 4a). It is unclear whether these mutations were related to the subsequent diagnosis and we therefore restricted our subsequent analysis on the risk of leukemias.

To ascertain the relationship between the mCA clone at initial blood draw and the clone at leukemic transformation, we analyzed leukemia samples from 16 patients that we also sequenced at the time of diagnosis by a comparable targeted panel (Methods). Among the 5 patients with detectable mCA, both gene mutations and mCAs coincided in the same clone that seeded the corresponding hematologic neoplasms (Fig. 4b and Supplementary Fig. 19). In all the cases these clones acquired additional alterations prior to disease transformation.

Aberrant blood counts are a common early symptom of hematologic neoplasms. However, cytopenias are frequent in cancer patients undergoing therapy. Formal hematological evaluations are typically delayed until cytopenias have proven to be persistent after therapy cessation. Serial blood counts in 4 of the cases that subsequently developed MDS (Fig. 4band Supplementary Fig. 19) showed that composite CH genotypes characteristic of leukemic transformation can be detected 9–39 months prior to diagnosis and before the onset of clinical-grade cytopenia43.

Implications for risk stratification

CH mutational features obtained by NGS assays are increasingly being interpreted in the clinic to evaluate a patient’s risk of leukemic transformation20,21. Established criteria are mainly based on gene mutations and involve PD status, variant allele frequency, and number of mutations10,11,22,23,26. Although it has been shown that certain mCA events pose an increased risk of subsequent blood cancers, it is unclear whether mCA status adds independent value to the stratification of patients at risk of disease progression when considered together with gene mutations13,17,18.

During patient follow-up, the 3-year cumulative incidence of leukemias (MDS, MPN, AML, CML, CLL) was significantly higher in patients with composite CH genotypes (14.6%, CI: 7–22%) as compared to patients with either mCA, gene mutation PD only, or neither, of which all had a 3-year cumulative incidence of less than 1% (Fig. 4c). We performed a multivariable cause-specific Cox regression analysis and showed that mCA was independently predictive of subsequent (>9 months post-sequencing) leukemia diagnosis (HR = 14, 95% CI: 6–33, P = 1.2e-09) after adjusting for number of mutations and VAF in PDs (Fig. 4d). Importantly, this effect is not restricted to the subset of patients that rapidly progress to overt disease. mCA remains a statistically significant predictor among solid tumor patients who survived and did not yet experience leukemic progression in the course of their clinical care at the time of different landmarks (3, 6, 9, and 12 months post-sequencing; Supplementary Fig. 20). In addition, we obtain similar risk estimates for gene mutations and mCA after restricting the analysis to patients with no blood abnormalities at the time of CH assessment (Supplementary Fig. 21). This demonstrates that incorporating features of chromosomal alterations into the clinical evaluation of CH has the potential to deliver better risk stratification as well as improve the diagnostic workup of patients with suspected hematologic disease.

Recent characterizations of somatic mutations in cancer as well as normal tissues suggest that the multi-step process of tumorigenesis is often an interplay between gene mutations and chromosomal alterations2,3,44. Studies of CH provide an opportunity to elucidate the mechanisms of mutation acquisition, clonal selection, and malignant transformation. Owing to limitations in genotyping platforms and sample size, most CH studies have largely been restricted to either gene mutations or mCAs (Supplementary Table 1)7,9,13,14,15,16,17,19. Consequently, such studies do not permit full characterization of the dynamic processes underlying the early steps of clonal evolution. Here, we present a large-scale integrative analysis of mCAs and gene mutations in CH. We leveraged data from a prospective sequencing cohort of 32,442 cancer patients profiled by high-depth targeted capture sequencing of 468 cancer genes and genome-wide copy number probes (MSK-IMPACT) to profile the landscape of mCAs in conjunction with gene mutations. Since our methodology is biased towards the detection of mCAs in relatively large clones (≥10%), it is likely that mCAs that are preferentially present in low cell fractions or in regions not optimally tiled by our panel may have been missed (Supplementary Figs. 5 and 6). Nonetheless, we were able to detect events with comparable sensitivity to prior studies based on unphased high-density SNP-arrays and recapitulate the genomic distribution, age-related incidence, and established demographical associations of mCAs14,15,16. In addition, our more sensitive approach to detect gene mutations (≥2% VAF) in the same sample relative to prior studies7,9,18,19 enables more comprehensive characterization of the co-mutational landscape of mCAs that we detect.

Although the selection forces driving the clonal expansion of somatic gene mutations are relatively well understood, the mechanisms underpinning the expansion of most mCAs remain unclear13,18,45. Recent studies have shown that at least 18% of CNLOH events can be attributed to germline alleles13,17,18. However, it is likely that the selection mechanisms of mCAs are multifactorial. Here we aim to uncover factors affecting the acquisition of mCAs in the context of demographic variables, environmental exposures, as well as CH gene mutations. We recapitulate known associations of mCA with age (OR = 1.8, P < 0.001), male gender (OR = 1.3, P = 0.012), and white race (OR = 1.5, P = 0.033)14,15,16. We find that mCAs are significantly associated with external beam radiation (OR = 1.7, P = 0.022), but not cytotoxic chemotherapy (OR = 0.9, P = 0.56). This result is in line with the highly heterogeneous effects of cancer therapy on CH shown in recent studies25,26. It is possible that associations may exist between specific types of mCAs and specific classes of cytotoxic therapy, but delineation of such effects will require larger datasets with more detailed treatment annotations.

In the present study, we analyze mCA presentation in the context of CH gene mutations and provide evidence that mCAs gain selective advantage through complex patterns of co-mutations. Approximately two-thirds of mCA cases had a concurrent gene mutation in one of the cancer-related genes in our panel. Out of the 57 coding gene mutations that co-localized with a deletion or CNLOH event in cis, 49 (86%) mapped in one of the 7 genes (DNMT3A, TET2, JAK2, MPL, EZH2, TP53, and ATM). This implies a gain of fitness that is mediated by bi-allelic alterations, resulting in either oncogene mutant dosage adjustment or inactivation of tumor suppressors. Certain mCA events, such as 7qCNLOH, 4/4qCNLOH, 9pCNLOH, and 1pCNLOH, were highly directed events acting on previously acquired gene mutations in the corresponding loci (EZH2, TET2, JAK2, and MPL). Together, these recurrent cis co-mutations underlied 26% of CNLOH events and 13% of all mCA events in this study. In addition, we observe recurrent composite genotypes indicative of co-operating or epistatic interactions as well as loss of gatekeeper function (i.e., TP53) presenting with multiple chromosomal aneuploidies. These co-mutations in trans underlied at least an additional 9.6% of mCA events. Taken together, our results reveal a previously under-characterized layer of complexity in the evolutionary dynamics of CH that converges towards characteristic genotypes associated with distinct leukemia subtypes. Analysis of paired samples at the time of transformation further demonstrated the co-presence of composite genotypes within the same precursor clone that further evolved to seed the subsequent leukemia. This puts mCAs in the context of the continuous evolutionary process of oncogenesis that can often span years and sheds lights on its patterns of acquisition and progression. Future developments of technology platforms that can reliably map gene mutations and chromosomal aberrations jointly at a single-cell resolution are warranted to fully resolve their clonal relationships.

Screening of CH mutations by NGS assays has been increasingly recognized as a potential strategy to identify individuals at risk of hematological neoplasms. This is particularly relevant for cancer patients, who are at a heightened risk of secondary leukemia. Existing screening approaches center around the gene-mutation features of CH including PD status, mutation number, and VAF. By multivariable analysis we show that mCA is an independent risk factor of leukemia in cancer patients. The mCA events that we characterize in this study are present in a relatively large cell fraction (≥10%), which means that they have already undergone substantial clonal expansion. It is likely that the selection forces and co-mutational patterns shaping the spectrum of smaller mCA clones are different. Similarly, interpretation of the clinical relevance of smaller mCA clones and their co-mutations will require more sensitive detection methods for subclonal alterations. It is also of note that due to the relatively short follow-up, we are limited in our capacity in assessing the long-term effects of mCAs on leukemia risk, which would be a goal for future studies. Nonetheless, our study demonstrates that the integration of chromosomal aberrations provides additional resolution to risk stratification as well as interpretation of clinical phenotypes that are often confounded by oncologic therapy (i.e., cytopenias), and that mCAs should be screened in conjunction with gene mutations. Beyond individuals at risk, our data support that there is an opportunity to also identify individuals with suspected hematologic disease that are difficult to capture with current clinical surveillance. To this end, we deliver a computational framework that enables the simultaneous characterization of mCA and gene mutation from a single assay. Fortuitously, solid tumor patients increasingly undergo routine sequencing of not only tumor biopsies, but also matched blood samples as germline controls28,46 thus enabling comprehensive CH surveillance programs to be readily integrated in routine cancer care.