Deep Learning Model for Tumor Type Prediction using Targeted Clinical Genomic Sequencing Data

Tumor type guides clinical treatment decisions in cancer, but histology-based diagnosis remains challenging. Genomic alterations are highly diagnostic of tumor type, and tumor type classifiers trained on genomic features have been explored, but the most accurate methods are not clinically feasible, relying on features derived from whole genome sequencing (WGS), or predicting across limited cancer types. We use genomic features from a dataset of 39,787 solid tumors sequenced using a clinical targeted cancer gene panel to develop Genome-Derived-Diagnosis Ensemble (GDD-ENS): a hyperparameter ensemble for classifying tumor type using deep neural networks. GDD-ENS achieves 93% accuracy for high-confidence predictions across 38 cancer types, rivalling performance of WGS-based methods. GDD-ENS can also guide diagnoses on rare type and cancers of unknown primary, and incorporate patient-specific clinical information for improved predictions. Overall, integrating GDD-ENS into prospective clinical sequencing workflows could provide clinically-relevant tumor type predictions to guide treatment decisions in real time.

Quantifying the Expanding Landscape of Clinical Actionability for Patients with Cancer

There is a continuing debate about the proportion of cancer patients that benefit from precision oncology, attributable in part to conflicting views as to which molecular alterations are clinically actionable. To quantify the expansion of clinical actionability since 2017, we annotated 47,271 solid tumors sequenced with the MSK-IMPACT clinical assay using two temporally distinct versions of the OncoKB knowledge base deployed 5 years apart. Between 2017 and 2022, we observed an increase from 8.9% to 31.6% in the fraction of tumors harboring a standard care (level 1 or 2) predictive biomarker of therapy response and an almost halving of tumors carrying nonactionable drivers (44.2% to 22.8%). In tumors with limited or no clinical actionability, TP53 (43.2%), KRAS (19.2%), and CDKN2A (12.2%) were the most frequently altered genes.

SIGNIFICANCE

Although clear progress has been made in expanding the availability of precision oncology-based treatment paradigms, our results suggest a continued unmet need for innovative therapeutic strategies, particularly for cancers with currently undruggable oncogenic drivers. See related commentary by Horak and Fröhling, p. 18. This article is featured in Selected Articles from This Issue, p. 5.

Deep Learning Model for Tumor Type Prediction using Targeted Clinical Genomic Sequencing Data

Tumor type guides clinical treatment decisions in cancer, but histology-based diagnosis remains challenging. Genomic alterations are highly diagnostic of tumor type, and tumor type classifiers trained on genomic features have been explored, but the most accurate methods are not clinically feasible, relying on features derived from whole genome sequencing (WGS), or predicting across limited cancer types. We use genomic features from a dataset of 39,787 solid tumors sequenced using a clinical targeted cancer gene panel to develop Genome-Derived-Diagnosis Ensemble (GDD-ENS): a hyperparameter ensemble for classifying tumor type using deep neural networks. GDD-ENS achieves 93% accuracy for high-confidence predictions across 38 cancer types, rivalling performance of WGS-based methods. GDD-ENS can also guide diagnoses on rare type and cancers of unknown primary, and incorporate patient-specific clinical information for improved predictions. Overall, integrating GDD-ENS into prospective clinical sequencing workflows has enabled clinically-relevant tumor type predictions to guide treatment decisions in real time.

Walking further. How surgery can help the cerebral palsy child

INTRODUCTION

The prevalence of cerebral palsy (CP) in Malaysia is estimated at 2.6 per 1000 live births which is comparable to that of Australian and European data with ranges of 2.3- 4.21,2. Surgical intervention for the improvement of gait function and mobility in CP is a common practice, however scarce literature of its outcomes is available in Southeast Asia. This paper aims to address and compare outcomes of surgical interventions in our centre with other countries.

MATERIAL AND METHODS

Patients with Spastic CP with Gross Motor Function Classification System (GMFCS) I-III that underwent lower limb surgical intervention in our centre from 2008-2018 were retrospectively reviewed for The Spinal Alignment and Range of Motion Measure ROM subscale (SAROMM) scores and Functional Mobility Scale (FMS) 18 months after surgery. Changes in SAROMM, FMS scores and minimal clinically important difference (MCID) were determined.

RESULTS

19 patients were included in the study with mean age of 12.58. All patients underwent muscle tendon procedures. Box plot analysis of SAROMM showed reduction of median scores at 6(26.3%) and 12(47.4%) months which plateaus at 18 months post-surgery. Repeated measure ANOVA analysis showed there was a statistically significant effect of time on SAROMM scores (p <0.001) with MCID of 13.4. Improvement of FMS scores was the most at 50m with 13 children (p < 0.05), one at 5m and five at 500m. None reported worsening of FMS scores at 18 months. There were no changes of GMFCS levels by the end of 18 months.

CONCLUSION

Surgeries performed on GMFCS I-III patients with the aim of gait improvement translates into improved mobility with results comparable to other countries.

Abstract 5440: Deep-learning model for tumor type classification enables enhanced clinical decision support in cancer diagnosis

Background: Knowledge of a patient’s tumor type is essential for guiding clinical treatment decisions in cancer, but histologically-based diagnosis remains challenging for a subset of cancers. Genomic alterations are highly indicative of tumor type and can be used to build classifiers that predict diagnoses, but most genomic-based classification methods use whole genome sequencing (WGS) data which is not feasible for widespread clinical implementation at present. Clinical sequencing is typically performed using cancer gene panels that target individual mutations, often drivers, but previous tumor type classifiers developed using driver-based features alone perform poorly. We hypothesize that a classifier developed using state-of-the-art deep-learning methods and a sufficiently large training cohort would be able to overcome previous accuracy limitations and support the development of a clinically-relevant tumor type prediction model.

Methods: We present Deep Genome-Derived Diagnosis (GDD-ENS), an ensemble-based deep-learning tumor type classification method trained using data from cancer gene panel sequencing. We specifically use data from MSK-IMPACT, an FDA-authorized clinical sequencing assay that reports genomic alterations including mutations, indels, copy number alterations, and gene fusions across 505 cancer-associated genes. We aggregated a discovery cohort of 35,372 patients with solid tumors profiled with MSK-IMPACT across 38 common cancer types and used this set to generate 4,487 somatic mutation features for development.

Results: GDD-ENS achieves 78.8% accuracy on a held out validation cohort of 6971 patients. For the 71.9% of predictions assigned a high confidence by the model, accuracy increases to 92.7%, rivaling WGS-based models. We use Shapley Values to report prediction-specific feature importance, and aggregate them across cancer types to show GDD-ENS identifies known cancer type-genomic alteration trends. GDD-ENS also, with high accuracy, identifies patients with cancer types not included in the 38 common types using metrics derived from ensemble statistics. For patients where non-genomic information could further guide predictions, we implement a customizable prediction-specific adaptive prior distribution and report improved accuracy after adjusting predictions to account for features such as metastatic sample biopsy site. Finally, we apply GDD-ENS to a set of 1,123 patients with Cancers of Unknown Primary (CUP) and return high confidence predictions for 49% of these patients, in some cases matching predictions on CUP samples with diagnoses that were later confirmed through additional sampling and disease progression.

Conclusions: Integrating GDD-ENS into prospective clinical sequencing workflows will enable clinically-relevant tumor type predictions that can guide treatment decisions in real-time.

Citation Format: Madison Darmofal, Shalabh Suman, Gurnit Atwal, Jie-Fu Chen, Anna Varghese, Jason C. Chang, Anoop Balakrishnan Rema, Aijazuddin Syed, Quaid Morris, Michael Berger. Deep-learning model for tumor type classification enables enhanced clinical decision support in cancer diagnosis. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5440.

Abstract 6585: OncoKB, MSK’s precision oncology knowledge base

OncoKB, Memorial Sloan Kettering Cancer Center’s (MSK) precision oncology knowledge base (www.oncokb.org), is an FDA-recognized* somatic variant database that contains information about the oncogenic effect and clinical implications of genomic alterations in cancer. Since its 2016 public release, OncoKB has grown to include annotation for &gt;5,770 alterations in ~700 cancer-associated genes. OncoKB data is integrated into the cBioPortal for Cancer Genomics and used to annotate &gt;12,000 MSK patient sequencing reports annually, encompassing both solid tumor and hematological malignancies. Users in academic, commercial and hospital settings outside MSK can programmatically access OncoKB data via its web API with an OncoKB license, which is free for academic research. To date, users from ~ 1400 institutions across &gt;70 countries have licensed access to OncoKB annotations. The OncoKB Therapeutic (Tx) Levels of Evidence assign tumor-type specific clinical actionability to individual mutational events based on data supporting whether an alteration is predictive of response to matched targeted therapies. To date, OncoKB includes 44 Level 1 genes (included in the FDA drug label), 23 Level 2 genes (included in professional guidelines), 33 Level 3A genes (predictive of drug response in well-powered clinical studies), 27 Level 4 genes (predictive of drug response based on compelling biological evidence), and 11 R1/R2 resistance genes. In 2022, several major content additions were made to OncoKB based on key shifts in the precision oncology landscape. For example, OncoKB included 2 new tumor-agnostic FDA drug approvals, dabrafenib + trametinib and selpercatinib for BRAF V600E and RET fusion-positive solid tumors respectively (Level 1), capturing 5 tumor-agnostic FDA drug approvals to date. OncoKB promoted ERBB2 oncogenic mutations and FGFR1 fusions to Level 1 following their inclusion as patient eligibility criteria in FDA drug labels for trastuzumab deruxtecan (NSCLC) and pemigatinib (myeloid/lymphoid neoplasms) respectively. NCCN guidelines for uterine sarcoma and pancreatic cancer listed PARP-inhibition for BRCA-mutant disease, making them Level 2 in these indications. Lastly, previously considered undruggable targets, TP53 Y220C and KRAS G12D, were included in OncoKB based on compelling evidence demonstrating response to allele-targeting drugs, PC14586 and RMC-6263, respectively. In sum, 7 novel clinically actionable biomarkers (Levels 1-4) and 11 follow-on precision oncology therapies for existing leveled biomarkers were added to OncoKB in 2022. Current OncoKB efforts are focused on prioritized high-volume cancer gene curation for annotation of whole exome/genome data, annotation of germline alterations and development of a clinical trials matching system. *FDA recognition of OncoKB is partial and limited to the information clearly marked on www.oncokb.org.

Citation Format: Sarah P. Suehnholz, Moriah Nissan, Hongxin Zhang, Ritika Kundra, Calvin Lu, Amanda Dhaneshwar, Nicole Fernandez, Stephanie Carrero, Maria E. Arcila, Marc Ladanyi, Michael F. Berger, Aijazuddin Syed, Rose Brannon, Ross Levine, Ahmet Dogan, Ezra Rosen, Alexander Drilon, David B. Solit, Nikolaus Schultz, Debyani Chakravarty. OncoKB, MSK’s precision oncology knowledge base. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 6585.

Diagnostic yield and clinical relevance of expanded genetic testing for cancer patients

BACKGROUND

Genetic testing (GT) for hereditary cancer predisposition is traditionally performed on selected genes based on established guidelines for each cancer type. Recently, expanded GT (eGT) using large hereditary cancer gene panels uncovered hereditary predisposition in a greater proportion of patients than previously anticipated. We sought to define the diagnostic yield of eGT and its clinical relevance in a broad cancer patient population over a 5-year period.

METHODS

A total of 17,523 cancer patients with a broad range of solid tumors, who received eGT at Memorial Sloan Kettering Cancer Center between July 2015 to April 2020, were included in the study. The patients were unselected for current GT criteria such as cancer type, age of onset, and/or family history of disease. The diagnostic yield of eGT was determined for each cancer type. For 9187 patients with five common cancer types frequently interrogated for hereditary predisposition (breast, colorectal, ovarian, pancreatic, and prostate cancer), the rate of pathogenic/likely pathogenic (P/LP) variants in genes that have been associated with each cancer type was analyzed. The clinical implications of additional findings in genes not known to be associated with a patients' cancer type were investigated.

RESULTS

16.7% of patients in a broad cancer cohort had P/LP variants in hereditary cancer predisposition genes identified by eGT. The diagnostic yield of eGT in patients with breast, colorectal, ovarian, pancreatic, and prostate cancer was 17.5%, 15.3%, 24.2%, 19.4%, and 15.9%, respectively. Additionally, 8% of the patients with five common cancers had P/LP variants in genes not known to be associated with the patient's current cancer type, with 0.8% of them having such a variant that confers a high risk for another cancer type. Analysis of clinical and family histories revealed that 74% of patients with variants in genes not associated with their current cancer type but which conferred a high risk for another cancer did not meet the current GT criteria for the genes harboring these variants. One or more variants of uncertain significance were identified in 57% of the patients.

CONCLUSIONS

Compared to targeted testing approaches, eGT can increase the yield of detection of hereditary cancer predisposition in patients with a range of tumors, allowing opportunities for enhanced surveillance and intervention. The benefits of performing eGT should be weighed against the added number of VUSs identified with this approach.

The context-specific role of germline pathogenicity in tumorigenesis

Human cancers arise from environmental, heritable and somatic factors, but how these mechanisms interact in tumorigenesis is poorly understood. Studying 17,152 prospectively sequenced patients with cancer, we identified pathogenic germline variants in cancer predisposition genes, and assessed their zygosity and co-occurring somatic alterations in the concomitant tumors. Two major routes to tumorigenesis were apparent. In carriers of pathogenic germline variants in high-penetrance genes (5.1% overall), lineage-dependent patterns of biallelic inactivation led to tumors exhibiting mechanism-specific somatic phenotypes and fewer additional somatic oncogenic drivers. Nevertheless, 27% of cancers in these patients, and most tumors in patients with pathogenic germline variants in lower-penetrance genes, lacked particular hallmarks of tumorigenesis associated with the germline allele. The dependence of tumors on pathogenic germline variants is variable and often dictated by both penetrance and lineage, a finding with implications for clinical management.

Enhanced specificity of clinical high-sensitivity tumor mutation profiling in cell-free DNA via paired normal sequencing using MSK-ACCESS

Circulating cell-free DNA from blood plasma of cancer patients can be used to non-invasively interrogate somatic tumor alterations. Here we develop MSK-ACCESS (Memorial Sloan Kettering - Analysis of Circulating cfDNA to Examine Somatic Status), an NGS assay for detection of very low frequency somatic alterations in 129 genes. Analytical validation demonstrated 92% sensitivity in de-novo mutation calling down to 0.5% allele frequency and 99% for a priori mutation profiling. To evaluate the performance of MSK-ACCESS, we report results from 681 prospective blood samples that underwent clinical analysis to guide patient management. Somatic alterations are detected in 73% of the samples, 56% of which have clinically actionable alterations. The utilization of matched normal sequencing allows retention of somatic alterations while removing over 10,000 germline and clonal hematopoiesis variants. Our experience illustrates the importance of analyzing matched normal samples when interpreting cfDNA results and highlights the importance of cfDNA as a genomic profiling source for cancer patients.

Liquid biopsies allow the non-invasive detection of somatic mutations from tumours. Here, the authors develop and test MSK-ACCESS, an NGS-based clinical assay for identifying low frequency mutations in 129 genes and describe how it benefits patients in the clinic.

155 Audit of Functional Outcome of Mason and Molloy Type 2 Posterior Malleolar Ankle Fractures Treated with Open Reduction and Internal Fixation Using A Posterolateral Approach

Introduction

It has been shown that direct fixation of the posterior malleolus improves functional outcomes. Our aim was to audit the functional outcome of patients with these fractures which were fixed with an isolated posterolateral approach.

Method

A consecutive case series of patients who underwent direct fixation of the posterior malleolus using a posterolateral approach between 20/12/2012 and 23/1/2020 was identified. Fractures were classified according to Mason and Molloy classification based on preoperative CT. Type 2a and 2b fractures were included. Functional outcome was assessed using Olerud-Molander score.

Result

18 patients were included. Mean age at time of surgery was 52 years (range 20 to 75 years). 56% (n = 10) were female. Mean follow up was 18.1 months (range 4.2 months to 7.2 years). OMAS score for type 2a fractures (n = 9) was 71.1 (95% CI 65.3 to 77.0). OMAS score for type 2b fractures (n = 9) was 67.8 (95% CI 54.6 to 81.0). There was no significant difference between groups (p = 0.65).

Conclusions

Fixation of Mason and Molloy Type 2 fractures using an isolated posterolateral approach results in satisfactory functional results for the majority of patients. Further prospective comparative study is needed to identify which patients benefit most from alternative approaches.

Chemical composition and synergistic effect of three Moroccan lavender EOs with ciprofloxacin against foodborne bacteria: a promising approach to modulate antimicrobial resistance

The aim of this study was to determine the chemical profile of the essential oils (EOs) of three Moroccan lavender species (Lavandula pedunculata, LP; Lavandula angustifolia, LA; and Lavandula maroccana, LM) and to investigate, for the first time, the synergistic effect of the optimal mixture of the EOs with conventional antibiotic ciprofloxacin against three pathogenic foodborne bacteria. Gas chromatography/mass spectrometry analysis showed that eucalyptol (39·05%), camphor (24·21%) and borneol (8·29%) were the dominant compounds of LA-EO. LP-EO was characterized by the abundance of camphor (74·51%) and fenchone (27·06%), whereas carvacrol (42·08%), camphor (17·95%) and fenchone (12·05%) were the main constituents of LM-EO. EOs alone or combined showed a remarkable antimicrobial activity against the tested bacteria with minimum inhibitory concentrations (MICs) ranging from 3·53 to 15·96 mg ml^-1 . The optimal mixture, calculated using a mixture design, corresponded to 19% LA, 38% LP and 43% LM. All combination of the EOs and the best EO mixture with ciprofloxacin exhibited a total synergism with fractional inhibitory concentration index values ranging from 0·27 to 0·37. The best EO mixture showed the highest gain of 128-fold, especially against Salmonella spp., more than that found testing the EOs separately. These findings should be taken into consideration for a possible application in the pharmaceutical and food industries.

Cancer therapy shapes the fitness landscape of clonal hematopoiesis

Acquired mutations are pervasive across normal tissues. However, understanding of the processes that drive transformation of certain clones to cancer is limited. Here we study this phenomenon in the context of clonal hematopoiesis (CH) and the development of therapy-related myeloid neoplasms (tMNs). We find that mutations are selected differentially based on exposures. Mutations in ASXL1 are enriched in current or former smokers, whereas cancer therapy with radiation, platinum and topoisomerase II inhibitors preferentially selects for mutations in DNA damage response genes (TP53, PPM1D, CHEK2). Sequential sampling provides definitive evidence that DNA damage response clones outcompete other clones when exposed to certain therapies. Among cases in which CH was previously detected, the CH mutation was present at tMN diagnosis. We identify the molecular characteristics of CH that increase risk of tMN. The increasing implementation of clinical sequencing at diagnosis provides an opportunity to identify patients at risk of tMN for prevention strategies.

Abstract 5703: Oncologic therapy shapes the fitness landscape of clonal hematopoiesis

Recent studies among healthy individuals show evidence of somatic mutations in leukemia-associated genes, referred to as clonal hematopoiesis (CH). To determine the relationship between CH and oncologic therapy we collected sequential blood samples from 525 cancer patients (median sampling interval time = 23 months, range: 6-53 months) of whom 61% received cytotoxic therapy or external beam radiation therapy and 39% received either targeted/immunotherapy or were untreated. Samples were sequenced using deep targeted capture-based platforms. To determine whether CH mutational features were associated with tMN risk, we performed Cox proportional hazards regression on 9,549 cancer patients exposed to oncologic therapy of whom 75 cases developed tMN (median time to transformation=26 months). To further compare the genetic and clonal relationships between tMN and the proceeding CH, we analyzed 35 cases for which paired samples were available. We compared the growth rate of the variant allele fraction (VAF) of CH clones across treatment modalities and in untreated patients. A significant increase in the growth rate of CH mutations was seen in DDR genes among those receiving cytotoxic (p=0.03) or radiation therapy (p=0.02) during the follow-up period compared to patients who did not receive therapy. Similar growth rates among treated and untreated patients were seen for non-DDR CH genes such as DNMT3A. Increasing cumulative exposure to cytotoxic therapy (p=0.01) and external beam radiation therapy (2x10-8) resulted in higher growth rates for DDR CH mutations. Among 34 subjects with at least two CH mutations in which one mutation was in a DDR gene and one in a non-DDR gene, we studied competing clonal dynamics for multiple gene mutations within the same patient. The risk of tMN was positively associated with CH in a known myeloid neoplasm driver mutation (HR=6.9, p&lt;10-6), and increased with the total number of mutations and clone size. The strongest associations were observed for mutations in TP53 and for CH with mutations in spliceosome genes (SRSF2, U2AF1 and SF3B1). Lower hemoglobin, lower platelet counts, lower neutrophil counts, higher red cell distribution width and higher mean corpuscular volume were all positively associated with increased tMN risk. Among 35 cases for which paired samples were available, in 19 patients (59%), we found evidence of at least one of these mutations at the time of pre-tMN sequencing and in 13 (41%), we identified two or more in the pre-tMN sample. In all cases the dominant clone at tMN transformation was defined by a mutation seen at CH Our serial sampling data provide clear evidence that oncologic therapy strongly selects for clones with mutations in the DDR genes and that these clones have limited competitive fitness, in the absence of cytotoxic or radiation therapy. We further validate the relevance of CH as a predictor and precursor of tMN in cancer patients. We show that CH mutations detected prior to tMN diagnosis were consistently part of the dominant clone at tMN diagnosis and demonstrate that oncologic therapy directly promotes clones with mutations in genes associated with chemo-resistant disease such as TP53.

Citation Format: Kelly L. Bolton, Ryan N. Ptashkin, Teng Gao, Lior Braunstein, Sean M. Devlin, Minal Patel, Antonin Berthon, Aijazuddin Syed, Mariko Yabe, Catherine Coombs, Nicole M. Caltabellotta, Mike Walsh, Ken Offit, Zsofia Stadler, Choonsik Lee, Paul Pharoah, Konrad H. Stopsack, Barbara Spitzer, Simon Mantha, James Fagin, Laura Boucai, Christopher J. Gibson, Benjamin Ebert, Andrew L. Young, Todd Druley, Koichi Takahashi, Nancy Gillis, Markus Ball, Eric Padron, David Hyman, Jose Baselga, Larry Norton, Stuart Gardos, Virginia Klimek, Howard Scher, Dean Bajorin, Eder Paraiso, Ryma Benayed, Maria Arcilla, Marc Ladanyi, David Solit, Michael Berger, Martin Tallman, Montserrat Garcia-Closas, Nilanjan Chatterjee, Luis Diaz, Ross Levine, Lindsay Morton, Ahmet Zehir, Elli Papaemmanuil. Oncologic therapy shapes the fitness landscape of clonal hematopoiesis [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5703.

Abstract PR08: Validation and clinical implementation of MSK-ACCESS, an ultra-deep sequencing assay for noninvasive somatic mutation profiling

Introduction: Circulating cell-free DNA (cfDNA) is a source of tumor-derived DNA to interrogate somatic alterations when tissue is not available or of insufficient quantity for analysis. At MSKCC, we have developed and validated MSK-ACCESS (Analysis of Circulating cfDNA to Evaluate Somatic Status), a targeted next-generation sequencing assay that can detect ultra-low frequency somatic variants in select exons and introns of 129 genes. MSK-ACCESS can identify mutations, copy number alterations, gene fusions, and MSI status in plasma and was recently approved by the NYS-DOH for clinical testing. Here, we present the results of the validation study and our clinical experience with MSK-ACCESS since June 2019.

Methods: Target regions from 129 genes were selected to maximize coverage of actionable, oncogenic, and hotspot mutations based on the first 25,000 tumors sequenced using MSK-IMPACT, our institutional clinical sequencing assay. Plasma cfDNA and buffy-coat DNA were extracted from whole blood collected in cell-stabilizing tubes (STRECK BCT cell-free DNA tube). Unique molecular indexes were introduced during DNA library construction, allowing for error suppression from consensus reads collapsed by Marianas, an in-house-developed algorithm. These consensus reads enable variant calling at low allelic frequency (AF) based on a 10−6 background error rate.

Results: Analytical validation of MSK-ACCESS demonstrated 93% accuracy (n=100 variants), 99% precision (n=153 variants), and 100% sensitivity based on an assay limit of detection of 0.5% AF (n=19 variants). Variants were detected down to 0.1% AF. To date, 240 clinical cfDNA and matched normal DNA pairs have been sequenced, analyzed for somatic alterations, and clinically reported to guide patient management. Most clinical cases were from lung (55%) or prostate (13%) cancers and submitted for diagnostic purposes (71%). Median raw coverage was 18,367X, and median consensus coverage was 1411X. Mutations were detected in 180 (75%) samples with a median variant AF of 1.8% (0.02% - 95%). Comparison of concurrent commercial plasma testing results to MSK-ACCESS revealed multiple variants that were of clonal hematopoiesis or germline origin incorrectly reported as somatic variants. In the lung cohort, 48 patients had tissue testing with MSK-IMPACT; among 32 patients with a driver alteration detected by MSK-ACCESS, 91% had the identical driver alteration reported by MSK-IMPACT. Additionally, MSK-ACCESS identified a MET exon 14 alteration in one lung cancer patient that led to protocol enrollment and partial response.

Conclusions: Liquid biopsy testing using MSK-ACCESS reliably detected clinically actionable mutations, reducing the need for multiple biopsies. These results also illustrate the importance of analyzing a matched normal sample when interpreting cfDNA results and highlight the potential of using cfDNA analysis to guide treatment selection, assess for treatment response, and identify mechanisms of treatment resistance.

This abstract is also being presented as Poster A20.

Citation Format: A. Rose Brannon, Gowtham Jayakumaran, Monica Diosdado, Yu Hu, Anna Razumova, Fanli Meng, Emily Lebow, Juber Patel, Ian Johnson, Preethi Srinivasan, Maysun Hasan, Jenna-marie Dix, Aijazuddin Syed, Brian Houck-Loomis, Bob T. Li, Charles Rudin, David Solit, Marc Ladanyi, Maria Arcila, Dana Tsui, Ahmet Zehir, Michael Berger, Ryma Benayed. Validation and clinical implementation of MSK-ACCESS, an ultra-deep sequencing assay for noninvasive somatic mutation profiling [abstract]. In: Proceedings of the AACR Special Conference on Advances in Liquid Biopsies; Jan 13-16, 2020; Miami, FL. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(11_Suppl):Abstract nr PR08.

The South Asian Health Foundation (UK) guidelines for managing diabetes during Ramadan

BACKGROUND

Fasting in the holy month of Ramadan is among the five pillars of Islam and is considered as a religious obligation by the Muslim population. People with diabetes observing the practice of fasts are at a higher risk of complications such as hypoglycaemia, hyperglycaemia and ketoacidosis due to changes in eating patterns and circadian rhythms. With the objective of mitigating these complications, the South Asian Health Foundation (UK) has developed the present guidelines based on robust evidence derived from epidemiological studies and clinical trials.

METHODS

We have highlighted the role of pre-Ramadan risk stratification and counselling by healthcare professionals with emphasis on the need for advice on adequate dietary and fluid intake, blood glucose monitoring and awareness of when to break the fast.

RESULTS

We reviewed the current literature and have given clinically-relevant recommendations on lifestyle modifications and glucose-lowering therapies such as metformin, sulphonylureas, dipeptidyl peptidase-4 inhibitors, sodium glucose co-transporter-2 inhibitors, thiazolidinediones, glucagon-like peptide-1 receptor agonists and insulin.

CONCLUSIONS

An individualised patient-centric treatment plan is essential to not only achieve optimal glycaemic outcomes but also enable people with diabetes to observe a risk-free month of fasting during Ramadan.

Development of Genome-Derived Tumor Type Prediction to Inform Clinical Cancer Care

IMPORTANCE

Diagnosing the site of origin for cancer is a pillar of disease classification that has directed clinical care for more than a century. Even in an era of precision oncologic practice, in which treatment is increasingly informed by the presence or absence of mutant genes responsible for cancer growth and progression, tumor origin remains a critical factor in tumor biologic characteristics and therapeutic sensitivity.

OBJECTIVE

To evaluate whether data derived from routine clinical DNA sequencing of tumors could complement conventional approaches to enable improved diagnostic accuracy.

DESIGN, SETTING, AND PARTICIPANTS

A machine learning approach was developed to predict tumor type from targeted panel DNA sequence data obtained at the point of care, incorporating both discrete molecular alterations and inferred features such as mutational signatures. This algorithm was trained on 7791 tumors representing 22 cancer types selected from a prospectively sequenced cohort of patients with advanced cancer.

RESULTS

The correct tumor type was predicted for 5748 of the 7791 patients (73.8%) in the training set as well as 8623 of 11 644 patients (74.1%) in an independent cohort. Predictions were assigned probabilities that reflected empirical accuracy, with 3388 cases (43.5%) representing high-confidence predictions (>95% probability). Informative molecular features and feature categories varied widely by tumor type. Genomic analysis of plasma cell-free DNA yielded accurate predictions in 45 of 60 cases (75.0%), suggesting that this approach may be applied in diverse clinical settings including as an adjunct to cancer screening. Likely tissues of origin were predicted from targeted tumor sequencing in 95 of 141 patients (67.4%) with cancers of unknown primary site. Applying this method prospectively to patients under active care enabled genome-directed reassessment of diagnosis in 2 patients initially presumed to have metastatic breast cancer, leading to the selection of more appropriate treatments, which elicited clinical responses.

CONCLUSIONS AND RELEVANCE

These results suggest that the application of artificial intelligence to predict tissue of origin in oncologic practice can act as a useful complement to conventional histologic review to provide integrated pathologic diagnoses, often with important therapeutic implications.

Communication practices for delivering health behaviour change conversations in primary care: a systematic review and thematic synthesis

BACKGROUND

Clinical guidelines exhort clinicians to encourage patients to improve their health behaviours. However, most offer little support on how to have these conversations in practice. Clinicians fear that health behaviour change talk will create interactional difficulties and discomfort for both clinician and patient. This review aims to identify how healthcare professionals can best communicate with patients about health behaviour change (HBC).

METHODS

We included studies which used conversation analysis or discourse analysis to study recorded interactions between healthcare professionals and patients. We followed an aggregative thematic synthesis approach. This involved line-by-line coding of the results and discussion sections of included studies, and the inductive development and hierarchical grouping of descriptive themes. Top-level themes were organised to reflect their conversational positioning.

RESULTS

Of the 17,562 studies identified through systematic searching, ten papers were included. Analysis resulted in 10 top-level descriptive themes grouped into three domains: initiating; carrying out; and closing health behaviour change talk. Of three methods of initiation, two facilitated further discussion, and one was associated with outright resistance. Of two methods of conducting behaviour change talk, one was associated with only minimal patient responses. One way of closing was identified, and patients did not seem to respond to this positively. Results demonstrated a series of specific conversational practices which clinicians use when talking about HBC, and how patients respond to these. Our results largely complemented clinical guidelines, providing further detail on how they can best be delivered in practice. However, one recommended practice - linking a patient's health concerns and their health behaviours - was shown to receive variable responses and to often generate resistance displays.

CONCLUSIONS

Health behaviour change talk is smoothly initiated, conducted, and terminated by clinicians and this rarely causes interactional difficulty. However, initiating conversations by linking a person's current health concern with their health behaviour can lead to resistance to advice, while other strategies such as capitalising on patient initiated discussions, or collaborating through question-answer sequences, may be well received.

Abstract 729: Zygosity, lineage, and penetrance dictate the role of germline pathogenicity in tumorigenesis

Human cancers arise from environmental, heritable, and somatic factors. While each has been studied extensively, how these factors interact to drive tumorigenesis and disease phenotypes is poorly understood. Here, we integrated germline pathogenicity with somatic alterations inferred from the prospective sequencing of 17,152 advanced cancer patients across 55 cancer types to explore their interplay in tumorigenesis. Leveraging expert curated germline variant calls from 5,358 patients, we developed a combined machine learning and risk stratification-based approach for pathogenicity assessment. Overall, 10.3% of patients harbored a pathogenic allele with penetrance levels ranging from high to low, and an additional 6.3% with a variant of uncertain penetrance. Integrating high-precision zygosity inference in the concomitant cancer diagnoses, we found that only carriers of high penetrance germline pathogenic alleles that are biallelically inactivated in their tumors had an earlier age of onset (51y in carriers vs. 58y in germline WT, p-value=3.5e-12), and an elevated rate of multiple independent cancer diagnoses (23% in carriers vs 13% in germline WT, p-value=1.6e-7). Broadly, penetrance dictated somatic dependence, with only those higher penetrant germline pathogenic alleles associated with lineage-dependent selective pressure for somatic biallelic inactivation (p-value&lt;2e-16). Tumor lineage dictated the substantial variability in the rate of tumor-specific biallelic inactivation of germline alleles, even among high penetrance genes. Among only high and moderate penetrance genes, high-risk cancers for which these alleles predispose and are therefore strongly associated by prevalence had the highest biallelic inactivation rates as compared to those for which no association exists (85 versus 30%, respectively; p-value&lt;2e-16). Consequently, nearly 30% of all tumors diagnosed in carriers of high penetrance germline alleles are likely sporadic cancers whose tumorigenesis is unrelated to the germline dysfunction. Moreover, tumorigenesis in only those carriers of high penetrance germline alleles biallelically inactivated in their tumors required fewer somatic oncogenic driver mutations overall to confer a selective growth advantage, emphasizing their distinct role as a founding event in disease pathogenesis. Collectively, the role of germline pathogenicity in tumorigenesis is determined by penetrance and zygosity in a manner that is lineage-dependent, thereby facilitating the discovery of true germline drivers and multiple distinct routes to tumorigenesis in affected patients, with implications for disease pathogenesis, screening, and ultimately therapy.

Citation Format: Chaitanya Bandlamudi, Preethi Srinivasan, Philip Jonsson, Yelena M. Kemel, Shweta S. Chavan, Allison L. Richards, Alex Penson, Craig M. Bielski, Chris Fong, Aijazuddin Syed, Gowtham Jayakumaran, Meera Prasad, Jason Hwee, Nikolaus Schultz, Semanti Mukherjee, Vijai Joseph, Diana Mandelkar, Ozge Birsoy, Liying Zhang, Jinru Shia, Ahmet Zehir, Marc Ladanyi, David M. Hyman, Kenneth Offit, Mark E. Robson, David B. Solit, Zsofia K. Stadler, Michael F. Berger, Barry S. Taylor. Zygosity, lineage, and penetrance dictate the role of germline pathogenicity in tumorigenesis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 729.

Abstract LB-304: Oncologic therapy for solid tumors alters the risk of clonal hematopoiesis

Solid tumor patients often suffer from cytopenias and are at risk for therapy-related myeloid neoplasms (tMN). Somatic mutations in leukemia-associated genes can occur in normal healthy individuals, referred to as clonal hematopoiesis (CH). CH is associated with cytopenias, risk of leukemia and cardiovascular disease. We and others have shown that CH is frequent in cancer patients. Characterization of the relationship between exposure to specific oncologic regimens and CH and how these relate to cytopenias and tMN risk would inform treatment decisions and tMN prevention strategies.

To determine the relationship between CH and oncologic therapy we interrogated CH in a cohort of 9045 solid tumor patients. Subjects were sequenced using a targeted panel of cancer-associated mutations used to screen tumor samples against a blood control sample. Mutation detection was performed on blood-derived sequencing data using the matched tumor as a comparator and accounted for background sequencing error rates.

CH was identified in 23% of patients. In multivariate regression analyses adjusted by age, CH was more often found in current smokers (OR=1.20, 95%CI=1.07-1.35, p&lt;0.001) and less often found in Asians compared to Whites (OR=0.72, 95%CI=0.56-0.89, p&lt;0.001). Smoking was associated with CH mutations in ASXL1 (OR=3.75, 95%CI=2.73-5.17, p&lt;0.001). There was a higher proportion of patients with CH among those who had received chemotherapy (OR=1.14, 95%CI=1.02-1.26, p=0.02) and those who had received external beam radiation therapy (OR=1.45, 95%CI=1.28-1.63, p&lt;0.001) prior to blood collection. Mutations in the DNA repair/cell cycle pathway (including TP53, PPM1D and CHEK2) were more common among patients who received chemotherapy and radiation therapy prior to IMPACT testing compared to those who were treatment naïve (p&lt;0.001). Exposure to prior cytotoxic chemotherapy (OR=1.20, 95%CI=1.02-1.30; p=0.007) and radiation therapy (OR=1.6, 95%CI=1.4-1.9, p&lt;0.001) was associated with having CH while exposure to immunotherapy and targeted therapy was not. Increasing cumulative dose of chemotherapy overall and cytotoxic therapy was associated with a higher likelihood of CH (p=0.015 and p=0.007 respectively). There was evidence of specific gene, treatment and dosage effects. To further examine the relationship between oncologic therapy and clonal evolution of CH, we collected 375 sequential samples at least 18 months apart. T mean change in VAF of CH mutations per year was found to increase by 0.17% in patients who did not receive further therapy during the follow-up time and 0.49% in those who were exposed to cytotoxic chemotherapy. A subset of patients with CH were consented to germline testing for cancer predisposition genes (N=6368). We observe a higher rate of CH among patients with a germline mutation in the cell cycle/DNA repair pathway (i.e. TP53) when compared to patients without germline mutations (OR=3.7, 95% CI: 1.35-9.35, p-value=0.01).

CH is frequent in solid tumor patients and can be reliably detected when a matched tumor normal targeted gene sequencing approach is performed. Beyond age, CH is strongly associated with race, smoking and importantly prior exposure to oncologic therapy with evidence of specific treatment effects. Screening of CH in cancer cohorts is critical to the development of future clinical guidelines and risk-adapted prevention strategies for tMN.

Note: This abstract was not presented at the meeting.

Citation Format: Kelly Bolton, Ryan Ptashkin, Lior Braunstein, Teng Gao, Sean M. Devlin, Daniel Kelly, Catherine Coombs, Minal Patel, Matahi Moarii, Elsa Bernard, Antonin Berthon, Laura Boucai, Dominik Glodzik, Axel Martin, Zsofia Stadler, Michael Walsh, Diana Mandelker, Akshar Patel, Jessica Schulman, Gunes Gundem, Aijazuddin Syed, Maria Arcila, David B. Solit, Mark E. Robson, Marc Ladanyi, Choonsik Lee, John Philip, Dean Bajorin, Montserrat Garcia-Closas, Stuart Gardos, David Hyman, Martin Tallman, Mariko Yabe, Kenneth Offit, Howard Scher, Virginia Klimek, Luis Diaz, Nilanjan Chatterjee, Michael F. Berger, Lindsay Morton, Ross Levine, Ahmet Zehir, Elli Papaemmanuil. Oncologic therapy for solid tumors alters the risk of clonal hematopoiesis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr LB-304.

Abstract 1672: Clinical validation of a genomics-based classifier to predict tissue of origin from targeted tumor sequencing

Introduction

Various classes of genomic alterations can now be robustly detected from targeted Next Generation Sequencing (NGS)-based tumor profiling assays. Despite recent advances in genome-directed cancer therapies and lineage-agnostic basket clinical trials, tissue of origin remains a critical determinant of tumor biology and therapeutic sensitivity. Therefore, efficiently harnessing the mutational information to predict tissue of origin from NGS data can further aid diagnosis and treatment.

Methods

We have developed a novel random forest machine learning classifier that infers tissue of origin from the mutational features of each tumor NGS profile. We trained and validated the model using prospective sequencing data from &gt;30,000 patients profiled by MSK-IMPACT, a custom FDA-authorized clinical sequencing assay. An initial version of this classifier trained on a smaller cohort, presented previously, has been further optimized to incorporate additional tumor types, genomic features, mutational signatures, and accessibility features for clinicians. Altogether, the model evaluates a comprehensive range of mutational features to generate predictions. In addition, we developed a framework for the prospective clinical implementation of our method that allows for extension to the expanding MSK-IMPACT cohort and utilization at the point of care.

Results

Overall, we predicted the correct cancer type in 74% of cases, with nearly half of cases predicted with high confidence (&gt;95%). In order to make this tool accessible to pathologists for real-time diagnostic and treatment decisions, we also have implemented APIs that transmit our classifier predictions to the cBioPortal for Cancer Genomics and the MPath console from MSK Molecular Diagnostics Service. With the additional samples and genomic features, our model predictions have been improved and are considered during clinical review and sign-out. This practice has brought about critical diagnostic changes and orthogonal validation in several cases. In one case, a patient referred to MSK with a misdiagnosis of metastatic breast cancer to the bladder underwent MSK-IMPACT sequencing for her primary breast tumor and bladder lesion. The model predictions, confirmed by orthogonal testing, classified two lesions as independent primaries of breast and bladder origins, leading to significant changes in treatment regimens.

Conclusion

Our work delineates the framework of utilizing and optimizing machine learning models on NGS-based sequencing data to aid diagnosis and treatment. The same framework can be applied to cell-free DNA sequencing, which will capture tumor spatial heterogeneity and improve classifier performance. These results indicate that leveraging machine learning to

predict tissue of origin complements conventional histologic review to provide integrated diagnoses, often with critical therapeutic implications.

Citation Format: Youyun Zheng, Alexander V. Penson, Niedzica Camacho, Evan Biederstedt, Ahmet Zehir, Cyriac Kandoth, Aijazuddin Syed, Anna M. Varghese, Hikmat A. Al-Ahmadie, Nikolaus Schultz, Marc Ladanyi, David B. Solit, David S. Klimstra, David Hyman, Barry Taylor, Michael F. Berger. Clinical validation of a genomics-based classifier to predict tissue of origin from targeted tumor sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1672.

Widespread Selection for Oncogenic Mutant Allele Imbalance in Cancer

Driver mutations in oncogenes encode proteins with gain-of-function properties that enhance fitness. Heterozygous mutations are thus viewed as sufficient for tumorigenesis. We describe widespread oncogenic mutant allele imbalance in 13,448 prospectively characterized cancers. Imbalance was selected for through modest dosage increases of gain-of-fitness mutations. Negative selection targeted haplo-essential effectors of the spliceosome. Loss of the normal allele comprised a distinct class of imbalance driven by competitive fitness, which correlated with enhanced response to targeted therapies. In many cancers, an antecedent oncogenic mutation drove evolutionarily dependent allele-specific imbalance. In other instances, oncogenic mutations co-opted independent copy-number changes via the evolutionary process of exaptation. Oncogenic allele imbalance is a pervasive evolutionary innovation that enhances fitness and modulates sensitivity to targeted therapy.

New developments in liquid-liquid extraction, surface modification and agglomerate-free processing of inorganic particles

This review describes new methods for the particle extraction through liquid-liquid interface (PELLI). The discovery of new surface modification techniques, advanced extractors and new adsorption mechanisms enabled novel applications of PELLI in nanotechnology of metals, quantum dots, oxides and hydroxides. Colloidal and interface chemistry of PELLI is emerging as a new area of technological and scientific interest. The progress achieved in the understanding of particle behavior and interactions at the liquid-liquid interface, phase transfer and interface reactions allowed for the development of new extraction mechanisms. An important breakthrough was the development of surface modification techniques for extraction of functional oxides. Especially important is the possibility of particle transfer from the synthesis medium to the device processing medium, which facilitates agglomerate-free processing of functional nanoparticles. Multifunctional extractor molecules were discovered and used as capping and reducing agents for particle synthesis or dispersing and charging agents for colloidal processing. The progress achieved in the development of extractors and extraction mechanisms has driven the advances in the surface modification and functionalization of materials. New PELLI techniques were used for the development of advanced materials and devices for optical, photovoltaic, energy storage, electronic, biomedical, sensor and other applications.

Validation of Memorial Sloan Kettering Cancer Center nomogram to detect non-sentinel lymph node metastases in a United Kingdom cohort

AIM

Axillary lymph node dissection, although associated with long-term morbidity, has been the standard of treatment for all nodepositive breast cancer patients. We assessed the risk prediction ability (validity) of Memorial Sloan Kettering Cancer Center (MSKCC) nomogram for non-sentinel lymph node metastases and analysed the outcome of patients with sentinel node metastases.

PATIENTS AND METHODS

All operable early breast cancer patients with sentinel node macro metastases (size > 2mm) who underwent axillary dissection from April 2009 to March 2015 were considered eligible. The risk of non-sentinel lymph node metastases was calculated using an online MSKCC calculator, and accuracy was determined based on the area under the receiver-operating characteristic curve (AUC-ROC). Tumour characteristics and overall survival were also analysed as secondary end points.

RESULTS

Of 1745 patients who were diagnosed with operable breast cancer during the study period, 114 patients were considered eligible. The AUC-ROC was 0.66 suggestive of lesser accuracy in prediction and not statistically significant (p value = 0.7303). Seventysix (50.7%) of these patients did not have any non-sentinel node metastases. At a mean follow up of four years, the disease-free survival was 86.4% and overall survival rate was 88.4%.

CONCLUSIONS

The MSKCC nomogram was unable to accurately predict the risk in our cohort of patients with more than half of this cohort of patients not requiring axillary dissection. These findings are consistent with other European studies. This study thus highlights the need for modified prediction model for European cohorts.

Clinical Sequencing Defines the Genomic Landscape of Metastatic Colorectal Cancer

Metastatic colorectal cancers (mCRCs) are clinically heterogeneous, but the genomic basis of this variability remains poorly understood. We performed prospective targeted sequencing of 1,134 CRCs. We identified splice alterations in intronic regions of APC and large in-frame deletions in CTNNB1, increasing oncogenic WNT pathway alterations to 96% of CRCs. Right-sided primary site in microsatellite stable mCRC was associated with shorter survival, older age at diagnosis, increased mutations, and enrichment of oncogenic alterations in KRAS, BRAF, PIK3CA, AKT1, RNF43, and SMAD4 compared with left-sided primaries. Left-sided tumors frequently had no identifiable genetic alteration in mitogenic signaling, but exhibited higher mitogenic ligand expression. Our results suggest different pathways to tumorigenesis in right- and left-sided microsatellite stable CRC that may underlie clinical differences.

Mutation Detection in Patients With Advanced Cancer by Universal Sequencing of Cancer-Related Genes in Tumor and Normal DNA vs Guideline-Based Germline Testing

IMPORTANCE

Guidelines for cancer genetic testing based on family history may miss clinically actionable genetic changes with established implications for cancer screening or prevention.

OBJECTIVE

To determine the proportion and potential clinical implications of inherited variants detected using simultaneous sequencing of the tumor and normal tissue ("tumor-normal sequencing") compared with genetic test results based on current guidelines.

DESIGN, SETTING, AND PARTICIPANTS

From January 2014 until May 2016 at Memorial Sloan Kettering Cancer Center, 10 336 patients consented to tumor DNA sequencing. Since May 2015, 1040 of these patients with advanced cancer were referred by their oncologists for germline analysis of 76 cancer predisposition genes. Patients with clinically actionable inherited mutations whose genetic test results would not have been predicted by published decision rules were identified. Follow-up for potential clinical implications of mutation detection was through May 2017.

EXPOSURE

Tumor and germline sequencing compared with the predicted yield of targeted germline sequencing based on clinical guidelines.

MAIN OUTCOMES AND MEASURES

Proportion of clinically actionable germline mutations detected by universal tumor-normal sequencing that would not have been detected by guideline-directed testing.

RESULTS

Of 1040 patients, the median age was 58 years (interquartile range, 50.5-66 years), 65.3% were male, and 81.3% had stage IV disease at the time of genomic analysis, with prostate, renal, pancreatic, breast, and colon cancer as the most common diagnoses. Of the 1040 patients, 182 (17.5%; 95% CI, 15.3%-19.9%) had clinically actionable mutations conferring cancer susceptibility, including 149 with moderate- to high-penetrance mutations; 101 patients tested (9.7%; 95% CI, 8.1%-11.7%) would not have had these mutations detected using clinical guidelines, including 65 with moderate- to high-penetrance mutations. Frequency of inherited mutations was related to case mix, stage, and founder mutations. Germline findings led to discussion or initiation of change to targeted therapy in 38 patients tested (3.7%) and predictive testing in the families of 13 individuals (1.3%), including 6 for whom genetic evaluation would not have been initiated by guideline-based testing.

CONCLUSIONS AND RELEVANCE

In this referral population with selected advanced cancers, universal sequencing of a broad panel of cancer-related genes in paired germline and tumor DNA samples was associated with increased detection of individuals with potentially clinically significant heritable mutations over the predicted yield of targeted germline testing based on current clinical guidelines. Knowledge of these additional mutations can help guide therapeutic and preventive interventions, but whether all of these interventions would improve outcomes for patients with cancer or their family members requires further study.

TRIAL REGISTRATION

clinicaltrials.gov Identifier: NCT01775072.

Therapy-Related Clonal Hematopoiesis in Patients with Non-hematologic Cancers Is Common and Associated with Adverse Clinical Outcomes

Clonal hematopoiesis (CH), as evidenced by recurrent somatic mutations in leukemia-associated genes, commonly occurs among aging human hematopoietic stem cells. We analyzed deep-coverage, targeted, next-generation sequencing (NGS) data of paired tumor and blood samples from 8,810 individuals to assess the frequency and clinical relevance of CH in patients with non-hematologic malignancies. We identified CH in 25% of cancer patients, with 4.5% harboring presumptive leukemia driver mutations (CH-PD). CH was associated with increased age, prior radiation therapy, and tobacco use. PPM1D and TP53 mutations were associated with prior exposure to chemotherapy. CH and CH-PD led to an increased incidence of subsequent hematologic cancers, and CH-PD was associated with shorter patient survival. These data suggest that CH occurs in an age-dependent manner and that specific perturbations can enhance fitness of clonal hematopoietic stem cells, which can impact outcome through progression to hematologic malignancies and through cell-non-autonomous effects on solid tumor biology.

Abstract 4380: Integrative genomics analysis of metastatic colorectal cancer

We performed an integrated clinical and bioinformatic analysis of colorectal cancers (CRCs) genotyped at our institution from 4/2014-7/2016 to comprehensively characterize genomic alterations in metastatic CRC (mCRC). We analyzed 1008 samples (474 primaries, 534 metastases) from 985 mCRC patients and 128 early stage CRCs sequenced with MSK-IMPACT, a hybridization capture next generation sequencing assay. Metastatic CRCs were divided into 3 groups by mutation burden and MSIsensor algorithm score: microsatellite stable (MSS) (n=939; 95%), microsatellite-high (MSI-H) (n=41; 4%), and ultra-mutated (n=5; 1%). Early stage CRC were enriched for MSI-H due to clinical selection and were 53% MSS, 44% MSI-H, and 4% ultra-mutated. Ultra-mutated tumors exhibited &gt;100 mutations and harbored hotspot mutations in POLE in 8 cases and a potential novel POLE alteration in 1 case.

We evaluated the frequency of oncogenic alterations in MSI-H and MSS mCRCs. The frequency of the resistance biomarkers KRAS and NRAS did not vary between MSI-H and MSS mCRCs (46% v 41%, p=0.6). Potentially actionable alterations were enriched in MSI-H tumors (78% v 33%, p&lt;0.001). Metastases did not have more actionable alterations than primary tumors. We classified clinically relevant targets using the OncoKB classification as level 2B (FDA-approved target in another disease type), 3A (target with compelling early clinical evidence in CRC), and 3B (target with compelling early clinical evidence in another disease type). Twelve percent (109/939) of MSS mCRCs had a level 2B target: BRAF V600E (5%), ERBB2 amplification (AMP) (4%), MET AMP (1%), BRCA1/BRCA2 alteration (1%), TSC1/TSC2 mutation (1%), EGFR mutation (&lt;1%), RET fusion (&lt;1%); there was significant enrichment of BRAF V600E (24%) and BRCA1/BRCA2 alterations (29%) in MSI-H versus MSS mCRC (p&lt;0.001). NTRK fusions, the main 3A alteration identified, occurred in 7% of MSI-H and &lt;1% MSS CRC (p&lt;0.01). Level 3B alterations at ≥ 1% in MSS CRC included PIK3CA (15%), NRAS (3%), AKT1 (1%), MAP2K1 (1%), and ERBB2 (1%) mutations and FGFR1 AMP (2%). PIK3CA and PTCH1 mutations were both enriched in MSI-H versus MSS mCRC (32% v 15%, p&lt;0.01; 27%,v &lt;1%, p&lt;0.001, respectively).

Analysis of mutation frequencies in 3 MSS CRC disease states - early stage resected primary (The Cancer Genome Atlas, TCGA), primary site of mCRC, and metastatic site - found significant depletion of FBXW7 mutations in metastases. We also found significant and progressive enrichment of TP53 alterations (58% TCGA, 73% primaries of mCRC, 79% metastases) and BRAF mutations (4% TCGA, 9% primaries of mCRC, 10% metastases) in advanced disease, suggesting a role of these genes in aggressive disease. One third of the BRAF mutations in our cohorts were not V600 but known to be oncogenic.

In this large dataset, we identified markers of advanced CRC and found that while MSI-H and MSS CRC have a similar frequency of resistance biomarkers, MSI-H CRC more commonly harbor actionable alterations.

Citation Format: Marla Lipsyc, Walid Chatila, Jaclyn F. Hechtman, Francisco Sanchez-Vega, Sumit Middha, Andrea Cercek, Zsofia Stadler, Ritika Kundra, Aijazuddin Syed, David M. Hyman, Ahmet Zehir, Armin Shahrokni, Anna Varghese, Diane Reidy, Neil H. Segal, Efsevia Vakiani, David B. Solit, Marc Ladanyi, Michael F. Berger, Nancy Kemeny, Leonard Saltz, Nikolaus Schultz, Rona Yaeger. Integrative genomics analysis of metastatic colorectal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 4380. doi:10.1158/1538-7445.AM2017-4380

Abstract 971: Genome directed diagnosis informs clinical cancer care

Cancer treatment is primarily guided by the organ of origin, which despite extensive histopathological and clinical evaluation, remains ambiguous in many cases and may be inaccurate in cases of occult primary. We determine the extent to which routine prospective tumour sequencing, combined with conventional histopathology, can be used to infer the tumor type and thereby direct clinical decisions.

Characteristic patterns of somatic mutations, broad and focal copy number alterations, structural rearrangements, mutational signatures, and other facets acquired from prospective tumor sequencing can inform the tissue of origin classification of patient disease. Such prospective sequencing of active cancer patients presents an opportunity to guide diagnosis and therapy beyond the identification of individual biomarkers of treatment response. Probabilistic classification allows for systematic combination of genome-directed diagnosis with conventional histopathology and clinical history in the course of disease management.

Using more than 10,000 tumors collected from advanced cancer patients at our institution and sequenced using a comprehensive cancer panel (MSK-IMPACT) encompassing 341 genes, we have developed a probabilistic classifier that infers the tumor type from nine broad categories of genomic aberrations. Scores from a RandomForest classifier are calibrated to correspond to the probability for each of the 22 interrogated tumor types. In 18% of all cases, a prediction is made with very high confidence (greater than 99%), of which just over 99% agree with the existing diagnosis. The classifier has been used to guide clinical decision making, including re-diagnoses that distinguish between a recurrence and a new primary, spare or encourage surgery, facilitate access to FDA-approved molecularly targeted drugs and prompt germline genetic testing.

We demonstrate that genome-directed diagnosis, based on alterations routinely identified from prospective sequencing, can inform clinical cancer care with the potential to improve patient outcomes.

Citation Format: Alexander Penson, Niedzica Camacho, Anna M. Varghese, Adam Abeshouse, Pedram Razavi, Aijazuddin Syed, Ahmet Zehir, Nikolaus Schultz, David B. Solit, David Hyman, Barry S. Taylor, Michael F. Berger. Genome directed diagnosis informs clinical cancer care [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 971. doi:10.1158/1538-7445.AM2017-971

Prospective Genomic Profiling of Prostate Cancer Across Disease States Reveals Germline and Somatic Alterations That May Affect Clinical Decision Making

PURPOSE

A long natural history and a predominant osseous pattern of metastatic spread are impediments to the adoption of precision medicine in patients with prostate cancer. To establish the feasibility of clinical genomic profiling in the disease, we performed targeted deep sequencing of tumor and normal DNA from patients with locoregional, metastatic non-castrate, and metastatic castration-resistant prostate cancer (CRPC).

METHODS

Patients consented to genomic analysis of their tumor and germline DNA. A hybridization capture-based clinical assay was employed to identify single nucleotide variations, small insertions and deletions, copy number alterations and structural rearrangements in over 300 cancer-related genes in tumors and matched normal blood.

RESULTS

We successfully sequenced 504 tumors from 451 patients with prostate cancer. Potentially actionable alterations were identified in DNA damage repair (DDR), PI3K, and MAP kinase pathways. 27% of patients harbored a germline or a somatic alteration in a DDR gene that may predict for response to PARP inhibition. Profiling of matched tumors from individual patients revealed that somatic TP53 and BRCA2 alterations arose early in tumors from patients who eventually developed metastatic disease. In contrast, comparative analysis across disease states revealed that APC alterations were enriched in metastatic tumors, while ATM alterations were specifically enriched in CRPC.

CONCLUSION

Through genomic profiling of prostate tumors representing the disease clinical spectrum, we identified a high frequency of potentially actionable alterations and possible drivers of disease initiation, metastasis and castration-resistance. Our findings support the routine use of tumor and germline DNA profiling for patients with advanced prostate cancer, for the purpose of guiding enrollment in targeted clinical trials and counseling families at increased risk of malignancy.

The relationship between eosinophilic esophagitis and esophageal cancer

Eosinophilic esophagitis (EoE) is a disorder characterized by long-term chronic eosinophilic inflammation, which may predispose patients for malignant transformation. To determine if EoE is associated with an increased risk for esophageal cancer (EC) over time. This was a cross-sectional population-based study using the Explorys Platform. Data were collected from January 1999 to December 2014 and the search terms 'Eosinophlic esophagitis', gastroesophageal reflux disease, and Barrett's esophagus were used. The latter two served as comparison groups. The term 'malignant tumor of esophagus' was searched in Explorys and was accepted only if it was established after the aforementioned diagnoses. Analysis was performed using social science statistical software. A total of 27,183,310 subjects were included in the study. Of those 5,370 (0.02%) had the diagnosis of EoE, 25,610 (0.09%) BE, and 2,606,640 (9.59%) gastroesophageal reflux (GERD). In general, diagnosis of EC was rare prior to age 40. Therefore, subsequent analysis was done on subjects between 40 to 79 years of age (N = 20,257,480). The prevalence of EC was 0.05% (N = 2420), in BE (N = 19,640) 2.65%, and in GERD (N = 1,696,230) 0.22%. No cases of EC were identified among those with EoE. We found no significant difference or increased risk in the rate of EC in the EoE group (P = 0.575) when compared to the control population. In this largest study, we found no association between esophageal cancer and EoE. Further long-term prospective studies are needed to better characterize the relationship between EoE and esophageal malignancy.

Determining if Positive Predictive Value using Laboratory Risk Indicator for Necrotising Fasciitis is Applicable in Malaysian Patients with Necrotising Fasciitis

Introduction: Necrotising fasciitis (NF) is a rapidly progressive infection of the subcutaneous tissue and fascia which spreads rapidly. The scoring system of Laboratory Risk Indicator for Necrotising Fasciitis (LRINEC) developed by Wong et al has been proposed as a tool for distinguishing NF and other soft tissue infections (STI) in Singapore. We set out to establish whether the LRINEC score is applicable in our Malaysian setting.

Materials and Methods: A cross sectional study of all patients admitted to our hospital diagnosed with NF or To Rule Out NF (TRO NF) between January 1st 2016 to 30th June 2016. The sensitivity, specificity, positive and negative predictive values were then calculated for LRINEC score of ≥ 6 and ≥ 8.

Results: Fourty-four patients were identified with the diagnosis of NF or TRO NF in the study. Twenty-seven patients (61.4%) were deemed post-operatively as having NF and 17 patients (38.6%) not having NF. A sensitivity of 59.3% and specificity of 47.1% when a LRINEC score of ≥ 6 was taken with positive predictive value (PPV) of 64.0% and the negative predictive value (NPV) of 42.1%. When score ≥ 8 was taken, the sensitivity was 48.1% and specificity of 58.8% with PPV of 65% and NPV of 41.7%.

Conclusion: The low sensitivity and low PPV achieved in this study as well as other studies makes the LRINEC score unsuitable to be used solely to distinguish NF with other soft tissue infections.

Variability in genomic alterations between right- and left-sided microsatellite stable (MSS) metastatic colorectal cancer and impact on survival

3534

Background: Metastatic colorectal cancers (mCRCs) with a right-sided primary site are associated with shorter survival and insensitivity to EGFR inhibitors compared to those originating in the left side of the colon or rectum. Methods: We performed targeted gene sequencing of 928 consecutive MSS mCRCs. Primary tumor site was divided into right-sided for cecum to distal transverse colon (n = 242), left-sided for splenic flexure to rectum (n = 673), or unknown colonic location (n = 13). Histologic subtypes were conventional (adenocarcinoma not otherwise specified), conventional with mucinous features ( < 50% mucinous), mucinous, signet ring, and poorly differentiated. To evaluate receptor tyrosine kinase (RTK) signaling, we analyzed ligand mRNA expression in TCGA. Results: Overall survival from time of metastasis was shorter for right-sided than left-sided primary site (survival at 5 years: 45% v 67%, p < 0.001). Right-sided tumors had more mutations (5.60 v 4.62 per MB, p < 0.001) but fewer copy-number changes (0.18 v 0.22 fraction genome altered, p = 0.001) compared to left-sided tumors. Alterations of KRAS, BRAF, PIK3CA, PTEN, AKT1, RNF43, and SMAD4 were significantly enriched in right-sided tumors, and of APC and TP53 in left-sided tumors. In a multivariate model, APC (HR = 0.7, p = 0.03), BRAF (HR = 3.7, p < 0.001), and KRAS (HR = 1.7, p < 0.01) alterations predicted for survival, but primary site did not (HR = 0.74, p = 0.07). Amphiregulin, epiregulin, neuregulin, and HGF expression was significantly higher in left-sided tumors. We found a higher proportion of conventional histology (83% v 63%) and moderate differentiation (82% v 69%) for left versus right-sided cases. Conclusions: We find that within MSS mCRC there are significantly more oncogenic mutations in right-sided tumors, and the difference in survival between right- and left-sided mCRC is primarily driven by differences in mutations. Left-sided tumors more commonly exhibit a “simpler” conventional histology that is lower grade and may rely on native RTK signaling, such as EGFR, for growth, providing a potential mechanism for the differential sensitivity to EGFR inhibitors seen by primary tumor site.

Comprehensive detection of germline variants by MSK-IMPACT, a clinical diagnostic platform for solid tumor molecular oncology and concurrent cancer predisposition testing

Background

The growing number of Next Generation Sequencing (NGS) tests is transforming the routine clinical diagnosis of hereditary cancers. Identifying whether a cancer is the result of an underlying disease-causing mutation in a cancer predisposition gene is not only diagnostic for a cancer predisposition syndrome, but also has significant clinical implications in the clinical management of patients and their families.

Methods

Here, we evaluated the performance of MSK-IMPACT (Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets) in detecting genetic alterations in 76 genes implicated in cancer predisposition syndromes. Output from hybridization-based capture was sequenced on an Illumina HiSeq 2500. A custom analysis pipeline was used to detect single nucleotide variants (SNVs), small insertions/deletions (indels) and copy number variants (CNVs).

Results

MSK-IMPACT detected all germline variants in a set of 233 unique patient DNA samples, previously confirmed by previous single gene testing. Reproducibility of variant calls was demonstrated using inter- and intra- run replicates. Moreover, in 16 samples, we identified additional pathogenic mutations other than those previously identified through a traditional gene-by-gene approach, including founder mutations in BRCA1, BRCA2, CHEK2 and APC, and truncating mutations in TP53, TSC2, ATM and VHL.

Conclusions

This study highlights the importance of the NGS-based gene panel testing approach in comprehensively identifying germline variants contributing to cancer predisposition and simultaneous detection of somatic and germline alterations.

Electronic supplementary material

The online version of this article (doi:10.1186/s12920-017-0271-4) contains supplementary material, which is available to authorized users.

Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients

Tumor molecular profiling is a fundamental component of precision oncology, enabling the identification of genomic alterations in genes and pathways that can be targeted therapeutically. The existence of recurrent targetable alterations across distinct histologically defined tumor types, coupled with an expanding portfolio of molecularly targeted therapies, demands flexible and comprehensive approaches to profile clinically relevant genes across the full spectrum of cancers. We established a large-scale, prospective clinical sequencing initiative using a comprehensive assay, MSK-IMPACT, through which we have compiled tumor and matched normal sequence data from a unique cohort of more than 10,000 patients with advanced cancer and available pathological and clinical annotations. Using these data, we identified clinically relevant somatic mutations, novel noncoding alterations, and mutational signatures that were shared by common and rare tumor types. Patients were enrolled on genomically matched clinical trials at a rate of 11%. To enable discovery of novel biomarkers and deeper investigation into rare alterations and tumor types, all results are publicly accessible.

Genetic Determinants of Cisplatin Resistance in Patients With Advanced Germ Cell Tumors

Purpose Owing to its exquisite chemotherapy sensitivity, most patients with metastatic germ cell tumors (GCTs) are cured with cisplatin-based chemotherapy. However, up to 30% of patients with advanced GCT exhibit cisplatin resistance, which requires intensive salvage treatment, and have a 50% risk of cancer-related death. To identify a genetic basis for cisplatin resistance, we performed whole-exome and targeted sequencing of cisplatin-sensitive and cisplatin-resistant GCTs. Methods Men with GCT who received a cisplatin-containing chemotherapy regimen and had available tumor tissue were eligible to participate in this study. Whole-exome sequencing or targeted exon-capture-based sequencing was performed on 180 tumors. Patients were categorized as cisplatin sensitive or cisplatin resistant by using a combination of postchemotherapy parameters, including serum tumor marker levels, radiology, and pathology at surgical resection of residual disease. Results TP53 alterations were present exclusively in cisplatin-resistant tumors and were particularly prevalent among primary mediastinal nonseminomas (72%). TP53 pathway alterations including MDM2 amplifications were more common among patients with adverse clinical features, categorized as poor risk according to the International Germ Cell Cancer Collaborative Group (IGCCCG) model. Despite this association, TP53 and MDM2 alterations predicted adverse prognosis independent of the IGCCCG model. Actionable alterations, including novel RAC1 mutations, were detected in 55% of cisplatin-resistant GCTs. Conclusion In GCT, TP53 and MDM2 alterations were associated with cisplatin resistance and inferior outcomes, independent of the IGCCCG model. The finding of frequent TP53 alterations among mediastinal primary nonseminomas may explain the more frequent chemoresistance observed with this tumor subtype. A substantial portion of cisplatin-resistant GCTs harbor actionable alterations, which might respond to targeted therapies. Genomic profiling of patients with advanced GCT could improve current risk stratification and identify novel therapeutic approaches for patients with cisplatin-resistant disease.