MSGene: a multistate model using genetic risk and the electronic health record applied to lifetime risk of coronary artery disease.

Coronary artery disease (CAD) is the leading cause of death among adults worldwide. Accurate risk stratification can support optimal lifetime prevention. Current methods lack the ability to incorporate new information throughout the life course or to combine innate genetic risk factors with acquired lifetime risk. We designed a general multistate model (MSGene) to estimate age-specific transitions across 10 cardiometabolic states, dependent on clinical covariates and a CAD polygenic risk score. This model is designed to handle longitudinal data over the lifetime to address this unmet need and support clinical decision-making. We analyze longitudinal data from 480,638 UK Biobank participants and compared predicted lifetime risk with the 30-year Framingham risk score. MSGene improves discrimination (C-index 0.71 vs 0.66), age of high-risk detection (C-index 0.73 vs 0.52), and overall prediction (RMSE 1.1% vs 10.9%), in held-out data. We also use MSGene to refine estimates of lifetime absolute risk reduction from statin initiation. Our findings underscore our multistate model's potential public health value for accurate lifetime CAD risk estimation using clinical factors and increasingly available genetics toward earlier more effective prevention.

Bayesian meta-analysis of penetrance for cancer risk.

Multi-gene panel testing allows many cancer susceptibility genes to be tested quickly at a lower cost making such testing accessible to a broader population. Thus, more patients carrying pathogenic germline mutations in various cancer-susceptibility genes are being identified. This creates a great opportunity, as well as an urgent need, to counsel these patients about appropriate risk-reducing management strategies. Counseling hinges on accurate estimates of age-specific risks of developing various cancers associated with mutations in a specific gene, ie, penetrance estimation. We propose a meta-analysis approach based on a Bayesian hierarchical random-effects model to obtain penetrance estimates by integrating studies reporting different types of risk measures (eg, penetrance, relative risk, odds ratio) while accounting for the associated uncertainties. After estimating posterior distributions of the parameters via a Markov chain Monte Carlo algorithm, we estimate penetrance and credible intervals. We investigate the proposed method and compare with an existing approach via simulations based on studies reporting risks for two moderate-risk breast cancer susceptibility genes, ATM and PALB2. Our proposed method is far superior in terms of coverage probability of credible intervals and mean square error of estimates. Finally, we apply our method to estimate the penetrance of breast cancer among carriers of pathogenic mutations in the ATM gene.

Bone marrow stromal cells induce chromatin remodeling in multiple myeloma cells leading to transcriptional changes.

The natural history of multiple myeloma is characterized by its localization to the bone marrow and its interaction with bone marrow stromal cells. The bone marrow stromal cells provide growth and survival signals, thereby promoting the development of drug resistance. Here, we show that the interaction between bone marrow stromal cells and myeloma cells (using human cell lines) induces chromatin remodeling of cis-regulatory elements and is associated with changes in the expression of genes involved in the cell migration and cytokine signaling. The expression of genes involved in these stromal interactions are observed in extramedullary disease in patients with myeloma and provides the rationale for survival of myeloma cells outside of the bone marrow microenvironment. Expression of these stromal interaction genes is also observed in a subset of patients with newly diagnosed myeloma and are akin to the transcriptional program of extramedullary disease. The presence of such adverse stromal interactions in newly diagnosed myeloma is associated with accelerated disease dissemination, predicts the early development of therapeutic resistance, and is of independent prognostic significance. These stromal cell induced transcriptomic and epigenomic changes both predict long-term outcomes and identify therapeutic targets in the tumor microenvironment for the development of novel therapeutic approaches.

Benefits and Harms of Interception and Early Detection of Cancer.

Strategies for early detection and interception of cancer are based on 2 synergistic elements: proactive search for asymptomatic cancer, precancer, or cancer predisposition and proactive disruption of cancer evolution. Benefits and harms of both these elements will vary widely depending on the screened populations, the types of cancers targeted, the detection modalities, and the health care delivery approaches following diagnosis. This article attempts to identify common elements that can inform the evaluation of alternative strategies across many of these scenarios.

Rejoinder to the discussion on "Bayesian meta-analysis of penetrance for cancer risk".

The five discussions of our paper provide several modeling alternatives, extensions, and generalizations that can potentially guide future research in meta-analysis. In this rejoinder, we briefly summarize and comment on some of those points.

Meta-analysis of breast cancer risk for individuals with PALB2 pathogenic variants.

Multigene panel testing now allows efficient testing of many cancer susceptibility genes leading to a larger number of mutation carriers being identified. They need to be counseled about their cancer risk conferred by the specific gene mutation. An important cancer susceptibility gene is PALB2. Multiple studies reported risk estimates for breast cancer (BC) conferred by pathogenic variants in PALB2. Due to the diverse modalities of reported risk estimates (age-specific risk, odds ratio, relative risk, and standardized incidence ratio) and effect sizes, a meta-analysis combining these estimates is necessary to accurately counsel patients with this mutation. However, this is not trivial due to heterogeneity of studies in terms of study design and risk measure. We utilized a recently proposed Bayesian random-effects meta-analysis method that can synthesize estimates from such heterogeneous studies. We applied this method to combine estimates from 12 studies on BC risk for carriers of pathogenic PALB2 mutations. The estimated overall (meta-analysis-based) risk of BC is 12.80% (6.11%-22.59%) by age 50 and 48.47% (36.05%-61.74%) by age 80. Pathogenic mutations in PALB2 makes women more susceptible to BC. Our risk estimates can help clinically manage patients carrying pathogenic variants in PALB2.

The Pancreatic Cancer Early Detection (PRECEDE) Study is a Global Effort to Drive Early Detection: Baseline Imaging Findings in High-Risk Individuals.

BACKGROUND: Pancreatic adenocarcinoma (PC) is a highly lethal malignancy with a survival rate of only 12%. Surveillance is recommended for high-risk individuals (HRIs), but it is not widely adopted. To address this unmet clinical need and drive early diagnosis research, we established the Pancreatic Cancer Early Detection (PRECEDE) Consortium. METHODS: PRECEDE is a multi-institutional international collaboration that has undertaken an observational prospective cohort study. Individuals (aged 18-90 years) are enrolled into 1 of 7 cohorts based on family history and pathogenic germline variant (PGV) status. From April 1, 2020, to November 21, 2022, a total of 3,402 participants were enrolled in 1 of 7 study cohorts, with 1,759 (51.7%) meeting criteria for the highest-risk cohort (Cohort 1). Cohort 1 HRIs underwent germline testing and pancreas imaging by MRI/MR-cholangiopancreatography or endoscopic ultrasound. RESULTS: A total of 1,400 participants in Cohort 1 (79.6%) had completed baseline imaging and were subclassified into 3 groups based on familial PC (FPC; n=670), a PGV and FPC (PGV+/FPC+; n=115), and a PGV with a pedigree that does not meet FPC criteria (PGV+/FPC-; n=615). One HRI was diagnosed with stage IIB PC on study entry, and 35.1% of HRIs harbored pancreatic cysts. Increasing age (odds ratio, 1.05; P<.001) and FPC group assignment (odds ratio, 1.57; P<.001; relative to PGV+/FPC-) were independent predictors of harboring a pancreatic cyst. CONCLUSIONS: PRECEDE provides infrastructure support to increase access to clinical surveillance for HRIs worldwide, while aiming to drive early PC detection advancements through longitudinal standardized clinical data, imaging, and biospecimen captures. Increased cyst prevalence in HRIs with FPC suggests that FPC may infer distinct biological processes. To enable the development of PC surveillance approaches better tailored to risk category, we recommend adoption of subclassification of HRIs into FPC, PGV+/FPC+, and PGV+/FPC- risk groups by surveillance protocols.

Optimal ensemble construction for multistudy prediction with applications to mortality estimation.

It is increasingly common to encounter prediction tasks in the biomedical sciences for which multiple datasets are available for model training. Common approaches such as pooling datasets before model fitting can produce poor out-of-study prediction performance when datasets are heterogeneous. Theoretical and applied work has shown multistudy ensembling to be a viable alternative that leverages the variability across datasets in a manner that promotes model generalizability. Multistudy ensembling uses a two-stage stacking strategy which fits study-specific models and estimates ensemble weights separately. This approach ignores, however, the ensemble properties at the model-fitting stage, potentially resulting in performance losses. Motivated by challenges in the estimation of COVID-attributable mortality, we propose optimal ensemble construction, an approach to multistudy stacking whereby we jointly estimate ensemble weights and parameters associated with study-specific models. We prove that limiting cases of our approach yield existing methods such as multistudy stacking and pooling datasets before model fitting. We propose an efficient block coordinate descent algorithm to optimize the loss function. We use our method to perform multicountry COVID-19 baseline mortality prediction. We show that when little data is available for a country before the onset of the pandemic, leveraging data from other countries can substantially improve prediction accuracy. We further compare and characterize the method's performance in data-driven simulations and other numerical experiments. Our method remains competitive with or outperforms multistudy stacking and other earlier methods in the COVID-19 data application and in a range of simulation settings.

Meta-Analysis of Breast Cancer Risk for Individuals with PALB2 Pathogenic Variants.

Background: Pathogenic variants in cancer susceptibility genes can now be tested efficiently and economically with the wide availability of multi-gene panel testing. This has resulted in an unprecedented rate of identifying individuals carrying pathogenic variants. These carriers need to be counselled about their future cancer risk conferred by the specific gene mutation. An important cancer susceptibility gene is PALB2. Several studies reported risk estimates for breast cancer (BC) associated with pathogenic variants in PALB2. Because of the variety of modalities (age specific risk, odds ratio, relative risk, and standardized incidence ratio) and effect sizes of these risk estimates, a meta-analysis of all of these estimates of BC risk is necessary to provide accurate counseling of patients with pathogenic variants in PALB2. The challenge, though, in combining these estimates is the heterogeneity of studies in terms of study design and risk measure. Methods: We utilized a recently proposed novel Bayesian random-effects meta-analysis method that can synthesize and combine information from such heterogeneous studies. We applied this method to combine estimates from twelve different studies on BC risk for carriers of pathogenic PALB2 mutations, out of which two report age-specific penetrance, one reports relative risk, and nine report odds ratios. Results: The estimated overall (meta-analysis based) risk of BC is 12.80% by age 50 (6.11%- 22.59%) and 48.47% by age 80 (36.05%-61.74%). Conclusion: Pathogenic mutations in PALB2 makes women more susceptible to BC. Our risk estimates can help clinically manage patients carrying pathogenic variants in PALB2.

MSGene: Derivation and validation of a multistate model for lifetime risk of coronary artery disease using genetic risk and the electronic health record.

Currently, coronary artery disease (CAD) is the leading cause of death among adults worldwide. Accurate risk stratification can support optimal lifetime prevention. We designed a novel and general multistate model (MSGene) to estimate age-specific transitions across 10 cardiometabolic states, dependent on clinical covariates and a CAD polygenic risk score. MSGene supports decision making about CAD prevention related to any of these states. We analyzed longitudinal data from 480,638 UK Biobank participants and compared predicted lifetime risk with the 30-year Framingham risk score. MSGene improved discrimination (C-index 0.71 vs 0.66), age of high-risk detection (C-index 0.73 vs 0.52), and overall prediction (RMSE 1.1% vs 10.9%), with external validation. We also used MSGene to refine estimates of lifetime absolute risk reduction from statin initiation. Our findings underscore the potential public health value of our novel multistate model for accurate lifetime CAD risk estimation using clinical factors and increasingly available genetics.

Bayesian combinatorial MultiStudy factor analysis.

Mutations in the BRCA1 and BRCA2 genes are known to be highly associated with breast cancer. Identifying both shared and unique transcript expression patterns in blood samples from these groups can shed insight into if and how the disease mechanisms differ among individuals by mutation status, but this is challenging in the high-dimensional setting. A recent method, Bayesian Multi-Study Factor Analysis (BMSFA), identifies latent factors common to all studies (or equivalently, groups) and latent factors specific to individual studies. However, BMSFA does not allow for factors shared by more than one but less than all studies. This is critical in our context, as we may expect some but not all signals to be shared by BRCA1-and BRCA2-mutation carriers but not necessarily other high-risk groups. We extend BMSFA by introducing a new method, Tetris, for Bayesian combinatorial multi-study factor analysis, which identifies latent factors that any combination of studies or groups can share. We model the subsets of studies that share latent factors with an Indian Buffet Process, and offer a way to summarize uncertainty in the sharing patterns using credible balls. We test our method with an extensive range of simulations, and showcase its utility not only in dimension reduction but also in covariance estimation. When applied to transcript expression data from high-risk families grouped by mutation status, Tetris reveals the features and pathways characterizing each group and the sharing patterns among them. Finally, we further extend Tetris to discover groupings of samples when group labels are not provided, which can elucidate additional structure in these data.

Clinical implications of conflicting variant interpretations in the cancer genetics clinic.

PURPOSE: The aim of this study was to describe the clinical impact of commercial laboratories issuing conflicting classifications of genetic variants. METHODS: Results from 2000 patients undergoing a multigene hereditary cancer panel by a single laboratory were analyzed. Clinically significant discrepancies between the laboratory-provided test reports and other major commercial laboratories were identified, including differences between pathogenic/likely pathogenic and variant of uncertain significance (VUS) classifications, via review of ClinVar archives. For patients carrying a VUS, clinical documentation was assessed for evidence of provider awareness of the conflict. RESULTS: Fifty of 975 (5.1%) patients with non-negative results carried a variant with a clinically significant conflict, 19 with a pathogenic/likely pathogenic variant reported in APC or MUTYH, and 31 with a VUS reported in CDKN2A, CHEK2, MLH1, MSH2, MUTYH, RAD51C, or TP53. Only 10 of 28 (36%) patients with a VUS with a clinically significant conflict had a documented discussion by a provider about the conflict. Discrepant counseling strategies were used for different patients with the same variant. Among patients with a CDKN2A variant or a monoallelic MUTYH variant, providers were significantly more likely to make recommendations based on the laboratory-reported classification. CONCLUSION: Our findings highlight the frequency of variant interpretation discrepancies and importance of clinician awareness. Guidance is needed on managing patients with discrepant variants to support accurate risk assessment.

Variant-specific Mendelian Risk Prediction Model.

Many pathogenic sequence variants (PSVs) have been associated with increased risk of cancers. Mendelian risk prediction models use Mendelian laws of inheritance to predict the probability of having a PSV based on family history, as well as specified PSV frequency and penetrance (agespecific probability of developing cancer given genotype). Most existing models assume penetrance is the same for any PSVs in a certain gene. However, for some genes (for example, BRCA1/2), cancer risk does vary by PSV. We propose an extension of Mendelian risk prediction models to relax the assumption that risk is the same for any PSVs in a certain gene by incorporating variant-specific penetrances and illustrating these extensions on two existing Mendelian risk prediction models, BRCAPRO and PanelPRO. Our proposed BRCAPRO-variant and PanelPRO-variant models incorporate variant-specific BRCA1/2 PSVs through the region classifications. Due to the sparsity of the variant information we classify BRCA1/2 PSVs into three regions; the breast cancer clustering region (BCCR), the ovarian cancer clustering region (OCCR), and an other region. Simulations were conducted to evaluate the performance of the proposed BRCAPRO-variant model compared to the existing BRCAPRO model which assumes the penetrance is the same for any PSVs in BRCA1 (and respectively BRCA2). Simulation results showed that the BRCAPRO-variant model was well calibrated to predict region-specific BRCA1/2 carrier status with high discrimination and accuracy on the region-specific level. In addition, we showed that the BRCAPRO-variant model achieved performance gains over the existing risk prediction models in terms of calibration without loss in discrimination and accuracy. We also evaluated the performance of the two proposed models, BRCAPRO-variant and PanelPRO-variant, on a cohort of 1,961 families from the Cancer Genetics Network (CGN). We showed that our proposed models provide region-specific PSV carrier probabilities with high accuracy, while the calibration, discrimination and accuracy of gene-specific PSV carrier probabilities were comparable to the existing gene-specific models. As more variant-specific PSV penetrances become available, we have shown that Mendelian risk prediction models can be extended to integrate the additional information, providing precise variant or region-specific PSV carrier probabilities and improving future cancer risk predictions.

Combining Breast Cancer Risk Prediction Models.

Accurate risk stratification is key to reducing cancer morbidity through targeted screening and preventative interventions. Multiple breast cancer risk prediction models are used in clinical practice, and often provide a range of different predictions for the same patient. Integrating information from different models may improve the accuracy of predictions, which would be valuable for both clinicians and patients. BRCAPRO is a widely used model that predicts breast cancer risk based on detailed family history information. A major limitation of this model is that it does not consider non-genetic risk factors. To address this limitation, we expand BRCAPRO by combining it with another popular existing model, BCRAT (i.e., Gail), which uses a largely complementary set of risk factors, most of them non-genetic. We consider two approaches for combining BRCAPRO and BCRAT: (1) modifying the penetrance (age-specific probability of developing cancer given genotype) functions in BRCAPRO using relative hazard estimates from BCRAT, and (2) training an ensemble model that takes BRCAPRO and BCRAT predictions as input. Using both simulated data and data from Newton-Wellesley Hospital and the Cancer Genetics Network, we show that the combination models are able to achieve performance gains over both BRCAPRO and BCRAT. In the Cancer Genetics Network cohort, we show that the proposed BRCAPRO + BCRAT penetrance modification model performs comparably to IBIS, an existing model that combines detailed family history with non-genetic risk factors.

High-dose melphalan treatment significantly increases mutational burden at relapse in multiple myeloma.

High-dose melphalan (HDM) improves progression-free survival in multiple myeloma (MM), yet melphalan is a DNA-damaging alkylating agent; therefore, we assessed its mutational effect on surviving myeloma cells by analyzing paired MM samples collected at diagnosis and relapse in the IFM 2009 study. We performed deep whole-genome sequencing on samples from 68 patients, 43 of whom were treated with RVD (lenalidomide, bortezomib, and dexamethasone) and 25 with RVD + HDM. Although the number of mutations was similar at diagnosis in both groups (7137 vs 7230; P = .67), the HDM group had significantly more mutations at relapse (9242 vs 13 383, P = .005). No change in the frequency of copy number alterations or structural variants was observed. The newly acquired mutations were typically associated with DNA damage and double-stranded breaks and were predominantly on the transcribed strand. A machine learning model, using this unique pattern, predicted patients who would receive HDM with high sensitivity, specificity, and positive prediction value. Clonal evolution analysis showed that all patients treated with HDM had clonal selection, whereas a static progression was observed with RVD. A significantly higher percentage of mutations were subclonal in the HDM cohort. Intriguingly, patients treated with HDM who achieved complete remission (CR) had significantly more mutations at relapse yet had similar survival rates as those treated with RVD who achieved CR. This similarity could have been due to HDM relapse samples having significantly more neoantigens. Overall, our study identifies increased genomic changes associated with HDM and provides rationale to further understand clonal complexity.

MyLynch: A Patient-Facing Clinical Decision Support Tool for Genetically-Guided Personalized Medicine in Lynch Syndrome.

Lynch syndrome (LS) is a hereditary cancer susceptibility condition associated with varying cancer risks depending on which of the five causative genes harbors a pathogenic variant; however, lifestyle and medical interventions provide options to lower those risks. We developed MyLynch, a patient-facing clinical decision support (CDS) web application that applies genetically-guided personalized medicine (GPM) for individuals with LS. The tool was developed in R Shiny through a patient-focused iterative design process. The knowledge base used to estimate patient-specific risk leveraged a rigorously curated literature review. MyLynch informs LS patients of their personal cancer risks, educates patients on relevant interventions, and provides patients with adjusted risk estimates, depending on the interventions they choose to pursue. MyLynch can improve risk communication between patients and providers while also encouraging communication among relatives with the goal of increasing cascade testing. As genetic panel testing becomes more widely available, GPM will play an increasingly important role in patient care, and CDS tools offer patients and providers tailored information to inform decision-making. MyLynch provides personalized cancer risk estimates and interventions to lower these risks for patients with LS.

In-depth analysis of alternative splicing landscape in multiple myeloma and potential role of dysregulated splicing factors.

Splicing changes are common in cancer and are associated with dysregulated splicing factors. Here, we analyzed RNA-seq data from 323 newly diagnosed multiple myeloma (MM) patients and described the alternative splicing (AS) landscape. We observed a large number of splicing pattern changes in MM cells compared to normal plasma cells (NPC). The most common events were alterations of mutually exclusive exons and exon skipping. Most of these events were observed in the absence of overall changes in gene expression and often impacted the coding potential of the alternatively spliced genes. To understand the molecular mechanisms driving frequent aberrant AS, we investigated 115 splicing factors (SFs) and associated them with the AS events in MM. We observed that ~40% of SFs were dysregulated in MM cells compared to NPC and found a significant enrichment of SRSF1, SRSF9, and PCB1 binding motifs around AS events. Importantly, SRSF1 overexpression was linked with shorter survival in two independent MM datasets and was correlated with the number of AS events, impacting tumor cell proliferation. Together with the observation that MM cells are vulnerable to splicing inhibition, our results may lay the foundation for developing new therapeutic strategies for MM. We have developed a web portal that allows custom alternative splicing event queries by using gene symbols and visualizes AS events in MM and subgroups. Our portals can be accessed at http://rconnect.dfci.harvard.edu/mmsplicing/ and https://rconnect.dfci.harvard.edu/mmleafcutter/ .

Hierarchical resampling for bagging in multistudy prediction with applications to human neurochemical sensing.

We propose the "study strap ensemble", which combines advantages of two common approaches to fitting prediction models when multiple training datasets ("studies") are available: pooling studies and fitting one model versus averaging predictions from multiple models each fit to individual studies. The study strap ensemble fits models to bootstrapped datasets, or "pseudo-studies." These are generated by resampling from multiple studies with a hierarchical resampling scheme that generalizes the randomized cluster bootstrap. The study strap is controlled by a tuning parameter that determines the proportion of observations to draw from each study. When the parameter is set to its lowest value, each pseudo-study is resampled from only a single study. When it is high, the study strap ignores the multi-study structure and generates pseudo-studies by merging the datasets and drawing observations like a standard bootstrap. We empirically show the optimal tuning value often lies in between, and prove that special cases of the study strap draw the merged dataset and the set of original studies as pseudo-studies. We extend the study strap approach with an ensemble weighting scheme that utilizes information in the distribution of the covariates of the test dataset. Our work is motivated by neuroscience experiments using real-time neurochemical sensing during awake behavior in humans. Current techniques to perform this kind of research require measurements from an electrode placed in the brain during awake neurosurgery and rely on prediction models to estimate neurotransmitter concentrations from the electrical measurements recorded by the electrode. These models are trained by combining multiple datasets that are collected in vitro under heterogeneous conditions in order to promote accuracy of the models when applied to data collected in the brain. A prevailing challenge is deciding how to combine studies or ensemble models trained on different studies to enhance model generalizability. Our methods produce marked improvements in simulations and in this application. All methods are available in the studyStrap CRAN package.

Loss of ribonuclease DIS3 hampers genome integrity in myeloma by disrupting DNA:RNA hybrid metabolism.

The ribonuclease DIS3 is one of the most frequently mutated genes in the hematological cancer multiple myeloma, yet the basis of its tumor suppressor function in this disease remains unclear. Herein, exploiting the TCGA dataset, we found that DIS3 plays a prominent role in the DNA damage response. DIS3 inactivation causes genomic instability by increasing mutational load, and a pervasive accumulation of DNA:RNA hybrids that induces genomic DNA double-strand breaks (DSBs). DNA:RNA hybrid accumulation also prevents binding of the homologous recombination (HR) machinery to double-strand breaks, hampering DSB repair. DIS3-inactivated cells become sensitive to PARP inhibitors, suggestive of a defect in homologous recombination repair. Accordingly, multiple myeloma patient cells mutated for DIS3 harbor an increased mutational burden and a pervasive overexpression of pro-inflammatory interferon, correlating with the accumulation of DNA:RNA hybrids. We propose DIS3 loss in myeloma to be a driving force for tumorigenesis via DNA:RNA hybrid-dependent enhanced genome instability and increased mutational rate. At the same time, DIS3 loss represents a liability that might be therapeutically exploited in patients whose cancer cells harbor DIS3 mutations.