Evaluating the Utility of 18F-FDG PET/CT in Cancer of Unknown Primary.

Cancer of unknown primary (CUP) represents a heterogeneous group of metastatic tumors for which standardized diagnostic work-up fails to identify the primary site. We aimed to describe the Peter MacCallum Cancer Centre experience with 18F-FDG PET/CT in extracervical CUP with respect to detection of a primary site and its impact on management. A secondary aim was to compare overall survival (OS) in patients with and without a detected primary site. Methods: CUP patients treated between 2014 and 2020 were identified from medical oncology clinics and 18F-FDG PET/CT records. Information collated from electronic medical records included the suspected primary site and treatment details before and after 18F-FDG PET/CT. Clinicopathologic details and genomic analysis were used to determine the clinically suspected primary site and compared against 2 independent masked reads of 18F-FDG PET/CT images by nuclear medicine specialists to determine sensitivity, specificity, accuracy, and the rate of detection of the primary site. Results: We identified 147 patients, 65% of whom had undergone molecular profiling. The median age at diagnosis was 61 y (range, 20-84 y), and the median follow-up time was 74 mo (range, 26-83 mo). Eighty-two percent were classified as having an unfavorable CUP subtype as per international guidelines.18F-FDG PET/CT demonstrated a primary site detection rate of 41%, resulted in a change in management in 22%, and identified previously occult disease sites in 37%. Median OS was 16.8 mo for all patients and 104.7 and 12.1 mo for favorable and unfavorable CUP subtypes, respectively (P < 0.0001). Median OS in CUP patients when using 18F-FDG PET/CT, clinicopathologic, and genomic information was 19.8 and 8.5 mo when a primary site was detected and not detected, respectively (P = 0.016). Multivariable analysis of survival adjusted for age and sex remained significant for identification of a potential primary site (P < 0.001), a favorable CUP (P < 0.001), and an Eastern Cooperative Oncology Group status of 1 or less (P < 0.001). Conclusion: 18F-FDG PET/CT plays a complementary role in CUP diagnostic work-up and was able to determine the likely primary site in 41% of cases. OS is improved with primary site identification, demonstrating the value of access to diagnostic 18F-FDG PET/CT for CUP patients.

Recent progress in molecular classification of phaeochromocytoma and paraganglioma.

Phaeochromocytomas (PC) and paragangliomas (PG) are neural crest cancers with high heritability. Recent advances in molecular profiling, including multi-omics and single cell genomics has identified up to seven distinct molecular subtypes. These subtypes are defined by mutations involving hypoxia-inducible factors (HIFs), Krebs cycle, kinase and WNT signalling, but are also defined by chromaffin differentiation states. PCPG have a dominant proangiogenic microenvironment linked to HIF pathway activity and are generally considered "immune cold" tumours with a high number of macrophages. PCPG subtypes can indicate increased metastatic risk but secondary mutations in telomere maintenance genes TERT or ATRX are required to drive the metastatic phenotype. Molecular profiling can identify molecular therapeutic (e.g. RET and EPAS1) and radiopharmaceutical targets while also helping to support variant pathogenicity and familial risk. Molecular profiling and subtyping of PCPG therefore confers the possibility of nuanced prognostication and individual treatment stratification but this still requires large-scale prospective validation.

The evolution of molecular management of carcinoma of unknown primary.

PURPOSE OF REVIEW: There is significant need to improve diagnostic and therapeutic options for patients with cancer of unknown primary (CUP). In this review, we discuss the evolving landscape of molecular profiling in CUP. RECENT FINDINGS: Molecular profiling is becoming accepted into the diagnostic work-up of CUP patients with tumour mutation profiling now described in international CUP guidelines. Although tissue-of-origin (ToO) molecular tests utilising gene-expression and DNA methylation have existed some time, their clinical benefit remains unclear. Novel technologies utilising whole genome sequencing and machine learning algorithms are showing promise in determining ToO, however further research is required prior to clinical application. A recent international clinical trial found patients treated with molecularly-guided therapy based on comprehensive-panel DNA sequencing had improved progression-free survival compared to chemotherapy alone, confirming utility of performing genomic profiling early in the patient journey. Small phase 2 trials have demonstrated that some CUP patients are responsive to immunotherapy, but the best way to select patients for treatment is not clear. SUMMARY: Management of CUP requires a multifaceted approach incorporating clinical, histopathological, radiological and molecular sequencing results to assist with identifying the likely ToO and clinically actionable genomic alternations. Rapidly identifying a subset of CUP patients who are likely to benefit from site specific therapy, targeted therapy and/or immunotherapy will improve patient outcomes.

Multi-omic analysis of SDHB-deficient pheochromocytomas and paragangliomas identifies metastasis and treatment-related molecular profiles.

Hereditary SDHB-mutant pheochromocytomas (PC) and paragangliomas (PG) are rare tumours with a high propensity to metastasize although their clinical behaviour is unpredictable. To characterize the genomic landscape of these tumours and identify metastasis biomarkers, we performed multi-omic analysis on 94 tumours from 79 patients using seven molecular methods. Sympathetic (chromaffin cell) and parasympathetic (non-chromaffin cell) PCPG had distinct molecular profiles reflecting their cell-of-origin and biochemical profile. TERT and ATRX-alterations were associated with metastatic PCPG and these tumours had an increased mutation load, and distinct transcriptional and telomeric features. Most PCPG had quiet genomes with some rare co-operative driver events observed, including EPAS1/HIF-2α mutations. Two mechanisms of acquired resistance to DNA alkylating chemotherapies were also detected - MGMT overexpression and mismatch repair-deficiency causing hypermutation. Our comprehensive multi-omic analysis of SDHB-mutant PCPG therefore identified features of metastatic disease and treatment response, expanding our understanding of these rare neuroendocrine tumours.

A genome-wide CRISPR/Cas9 screen identifies DNA-PK as a sensitiser to 177Lutetium-DOTA-octreotate radionuclide therapy.

Peptide receptor radionuclide therapy (PRRT) using 177Lutetium-DOTA-octreotate (LuTate) for neuroendocrine tumours (NET) is now an approved treatment available in many countries, though primary or secondary resistance continue to limit its effectiveness or durability. We hypothesised that a genome-wide CRISPR/Cas9 screen would identify key mediators of response to LuTate and gene targets that might offer opportunities for novel combination therapies for NET patients. Methods: We utilised a genome-wide CRISPR-Cas9 screen in LuTate-treated cells to identify genes that impact on the sensitivity or resistance of cells to LuTate. Hits were validated through single-gene knockout. LuTate-resistant cells were assessed to confirm LuTate uptake and retention, and persistence of somatostatin receptor 2 (SSTR2) expression. Gene knockouts conferring LuTate sensitivity were further characterised by pharmacological sensitisation using specific inhibitors and in vivo analysis of the efficacy of these inhibitors in combination with LuTate. Results: The CRISPR-Cas9 screen identified several potential targets for both resistance and sensitivity to PRRT. Two gene knockouts which conferred LuTate resistance in vitro, ARRB2 and MVP, have potential mechanisms related to LuTate binding and retention, and modulation of DNA-damage repair (DDR) pathways, respectively. The screen showed that sensitivity to LuTate treatment in vitro can be conferred by the loss of a variety of genes involved in DDR pathways, with loss of genes involved in Non-Homologous End-Joining (NHEJ) being the most lethal. Loss of the key NHEJ gene, PRKDC (DNA-PK), either by gene loss or inhibition by two different inhibitors, resulted in significantly reduced cell survival upon exposure of cells to LuTate. In SSTR2-positive xenograft-bearing mice, the combination of nedisertib (a DNA-PK specific inhibitor) and LuTate produced a more robust control of tumour growth and increased survival compared to LuTate alone. Conclusions: DDR pathways are critical for sensing and repairing radiation-induced DNA damage, and our study shows that regulation of DDR pathways may be involved in both resistance and sensitivity to PRRT. Additionally, the use of a DNA-PK inhibitor in combination with LuTate PRRT significantly improves the efficacy of the treatment in pre-clinical models, providing further evidence for the clinical efficacy of this combination.

Six-year experience of Australia's first dedicated cancer of unknown primary clinic.

BACKGROUND: Diagnosis and management of cancers of unknown primary (CUP) remain challenging. This study examines the referral patterns, management and outcomes of patients referred to Australia's first dedicated CUP clinic. METHODS: Retrospective medical record review was conducted for patients seen at the Peter MacCallum Cancer Centre CUP clinic between July 2014 and August 2020. Overall survival (OS) was analysed for patients with a CUP diagnosis where treatment information was available. RESULTS: Of 361 patients referred, fewer than half had completed diagnostic work-up at the time of referral. A diagnosis of CUP was established in 137 (38%), malignancy other than CUP in 177 (49%) and benign pathology in 36 (10%) patients. Genomic testing was successfully completed in 62% of patients with initial provisional CUP and impacted management in 32% by identifying a tissue of origin or actionable genomic alteration. The use of site-specific, targeted therapy or immunotherapy was independently associated with longer OS compared to empirical chemotherapy. CONCLUSION: Our specialised CUP clinic facilitated diagnostic work-up among patients with suspected malignancy and provided access to genomic testing and clinical trials for patients with a CUP diagnosis, all of which are important to improve outcomes in this patient population.

Integrating Functional Imaging and Molecular Profiling for Optimal Treatment Selection in Neuroendocrine Neoplasms (NEN).

PURPOSE OF REVIEW: Gastroenteropancreatic NEN (GEP-NEN) are group of malignancies with significant clinical, anatomical and molecular heterogeneity. High-grade GEP-NEN in particular present unique management challenges. RECENT FINDINGS: In the current era, multidisciplinary management with access to a combination of functional imaging and targeted molecular profiling can provide important disease characterisation, guide individualised management and improve patient outcome. Multiple treatment options are now available, and combination and novel therapies are being explored in clinical trials. Precision medicine is highly relevant for a heterogenous disease like NEN. The integration of dual-tracer functional PET/CT imaging, molecular histopathology and genomic data has the potential to be used to gain a more comprehensive understanding of an individual patient's disease biology for precision diagnosis, prognostication and optimal treatment allocation.

A qualitative study of patients with Cancer of Unknown Primary: Perceptions of communication, understanding of diagnosis and genomic testing, and information needs.

OBJECTIVE: Patients with Cancer of Unknown Primary (CUP) commonly report poor understanding of their illness and high levels of psychological distress. Despite the potential benefits to CUP patients, there is a paucity of research exploring the reasons behind poor understanding of a CUP diagnosis. The aim of this study was to understand patients' experiences of communication with doctors, their understanding of diagnosis and the role of genomic testing, as well as their information needs. METHODS: Semi-structured interviews explored CUP patients' perceptions of communication with their doctors, understanding of their illness, and their needs regarding medical information. Qualitative inductive thematic analysis of transcribed audio-recordings was employed. SETTING/PARTICIPANTS: Nineteen patients were recruited from within a prospective cohort study involving routine genomic testing of CUP patients. RESULTS: CUP patients had varied perceptions of communication with doctors as well as different levels of need, readiness, and capacity for information. Some patients felt well understood and supported by their doctors while others did not. Many patients reported feeling overwhelmed and shocked when receiving their cancer diagnosis and emphasized the importance of family support in receiving and understanding medical information. While patients understood the implications of genomic testing for treatment and diagnosis, few had a detailed understanding of genomic testing. CONCLUSIONS: Patients' experience of communication and understanding of CUP could be potentially improved by clinicians' assessment of the communication style preferred by each patient and their family and the development of online resources to meet their evolving information needs.

A comparison of DNA sequencing and gene expression profiling to assist tissue of origin diagnosis in cancer of unknown primary.

Cancer of unknown primary (CUP) is a syndrome defined by clinical absence of a primary cancer after standardised investigations. Gene expression profiling (GEP) and DNA sequencing have been used to predict primary tissue of origin (TOO) in CUP and find molecularly guided treatments; however, a detailed comparison of the diagnostic yield from these two tests has not been described. Here, we compared the diagnostic utility of RNA and DNA tests in 215 CUP patients (82% received both tests) in a prospective Australian study. Based on retrospective assessment of clinicopathological data, 77% (166/215) of CUPs had insufficient evidence to support TOO diagnosis (clinicopathology unresolved). The remainder had either a latent primary diagnosis (10%) or clinicopathological evidence to support a likely TOO diagnosis (13%) (clinicopathology resolved). We applied a microarray (CUPGuide) or custom NanoString 18-class GEP test to 191 CUPs with an accuracy of 91.5% in known metastatic cancers for high-medium confidence predictions. Classification performance was similar in clinicopathology-resolved CUPs - 80% had high-medium predictions and 94% were concordant with pathology. Notably, only 56% of the clinicopathology-unresolved CUPs had high-medium confidence GEP predictions. Diagnostic DNA features were interrogated in 201 CUP tumours guided by the cancer type specificity of mutations observed across 22 cancer types from the AACR Project GENIE database (77,058 tumours) as well as mutational signatures (e.g. smoking). Among the clinicopathology-unresolved CUPs, mutations and mutational signatures provided additional diagnostic evidence in 31% of cases. GEP classification was useful in only 13% of cases and oncoviral detection in 4%. Among CUPs where genomics informed TOO, lung and biliary cancers were the most frequently identified types, while kidney tumours were another identifiable subset. In conclusion, DNA and RNA profiling supported an unconfirmed TOO diagnosis in one-third of CUPs otherwise unresolved by clinicopathology assessment alone. DNA mutation profiling was the more diagnostically informative assay. © 2022 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Rapid Point-of-Care Electrochemical Sensor for the Detection of Cancer Tn Antigen Carbohydrate in Whole Unprocessed Blood.

Improving outcomes for cancer patients during treatment and monitoring for cancer recurrence requires personalized care which can only be achieved through regular surveillance for biomarkers. Unfortunately, routine detection for blood-based biomarkers is cost-prohibitive using currently specialized laboratories. Using a rapid self-assembly sensing interface amenable to methods of mass production, we demonstrate the ability to detect and quantify a small carbohydrate-based cancer biomarker, Tn antigen (αGalNAc-Ser/Thr) in a small volume of blood, using a test format strip reminiscent of a blood glucose test. The detection of Tn antigen at picomolar levels is achieved through a new transduction mechanism based on the impact of Tn antigen interactions on the molecular dynamic motion of a lectin cross-linked lubricin antifouling brush. In tests performed on retrospective blood plasma samples from patients presenting three different tumor types, differentiation between healthy and diseased patients was achieved, highlighting the clinical potential for cancer monitoring.

Single-nuclei and bulk-tissue gene-expression analysis of pheochromocytoma and paraganglioma links disease subtypes with tumor microenvironment.

Pheochromocytomas (PC) and paragangliomas (PG) are rare neuroendocrine tumors associated with autonomic nerves. Here we use single-nuclei RNA-seq and bulk-tissue gene-expression data to characterize the cellular composition of PCPG and normal adrenal tissues, refine tumor gene-expression subtypes and make clinical and genotypic associations. We confirm seven PCPG gene-expression subtypes with significant genotype and clinical associations. Tumors with mutations in VHL, SDH-encoding genes (SDHx) or MAML3-fusions are characterized by hypoxia-inducible factor signaling and neoangiogenesis. PCPG have few infiltrating lymphocytes but abundant macrophages. While neoplastic cells transcriptionally resemble mature chromaffin cells, early chromaffin and neuroblast markers are also features of some PCPG subtypes. The gene-expression profile of metastatic SDHx-related PCPG indicates these tumors have elevated cellular proliferation and a lower number of non-neoplastic Schwann-cell-like cells, while GPR139 is a potential theranostic target. Our findings therefore clarify the diverse transcriptional programs and cellular composition of PCPG and identify biomarkers of potential clinical significance.

Psychological distress, understanding of cancer and illness uncertainty in patients with Cancer of Unknown Primary.

OBJECTIVE: Patients diagnosed with Cancer of Unknown Primary (CUP) experience high levels of psychological distress and report poor understanding of their cancer. We aimed to investigate: (1) if CUP patients with poorer understanding of their cancer diagnosis and testing experience more symptoms of psychological distress than those with better understanding; (2) if the relationship between patients' understanding of their cancer and psychological distress is mediated by illness uncertainty; and (3) explore whether patients' degree of understanding of their cancer can be predicted by clinical and socio-demographic factors. METHODS: 209 CUP patients completed a questionnaire measuring anxiety, depression, illness uncertainty, fatigue, pain, sleep and understanding of their cancer. Using an apriori theoretical framework, we employed structural equation modelling to investigate predictors of patient's understanding of their cancer and psychological distress and the relationships between understanding, illness uncertainty and distress. RESULTS: The structural equation model displayed good fit indices and supported the hypothesised relationship of patient's understanding of their cancer and the extent of psychological distress, which was mediated via illness uncertainty. Physical symptoms were positively associated with psychological distress and illness uncertainty. Younger age was predictive of lower patient's understanding of their cancer and higher levels of psychological distress. CONCLUSIONS: Patients with CUP, particularly those who are younger and experiencing more physical symptoms, report higher levels of psychological distress and may require additional mental health support. Our findings highlight a need to improve CUP patient's understanding about their illness, which could help reduce their illness uncertainty and alleviate psychological distress.

Molecular and genomic characterisation of a panel of human anal cancer cell lines.

Anal cancer is a rare disease that has doubled in incidence over the last four decades. Current treatment and survival of patients with this disease has not changed substantially over this period of time, due, in part, to a paucity of preclinical models to assess new therapeutic options. To address this hiatus, we set-out to establish, validate and characterise a panel of human anal squamous cell carcinoma (ASCC) cell lines by employing an explant technique using fresh human ASCC tumour tissue. The panel of five human ASCC cell lines were validated to confirm their origin, squamous features and tumourigenicity, followed by molecular and genomic (whole-exome sequencing) characterisation. This panel recapitulates the genetic and molecular characteristics previously described in ASCC including phosphoinositide-3-kinase (PI3K) mutations in three of the human papillomavirus (HPV) positive lines and TP53 mutations in the HPV negative line. The cell lines demonstrate the ability to form tumouroids and retain their tumourigenic potential upon xenotransplantation, with varied inducible expression of major histocompatibility complex class I (MHC class I) and Programmed cell death ligand 1 (PD-L1). We observed differential responses to standard chemotherapy, radiotherapy and a PI3K specific molecular targeted agent in vitro, which correlated with the clinical response of the patient tumours from which they were derived. We anticipate this novel panel of human ASCC cell lines will form a valuable resource for future studies into the biology and therapeutics of this rare disease.

γδ T Cells in Merkel Cell Carcinomas Have a Proinflammatory Profile Prognostic of Patient Survival.

Merkel cell carcinomas (MCC) are immunogenic skin cancers associated with viral infection or UV mutagenesis. To study T-cell infiltrates in MCC, we analyzed 58 MCC lesions from 39 patients using multiplex-IHC/immunofluorescence (m-IHC/IF). CD4+ or CD8+ T cells comprised the majority of infiltrating T lymphocytes in most tumors. However, almost half of the tumors harbored prominent CD4/CD8 double-negative (DN) T-cell infiltrates (>20% DN T cells), and in 12% of cases, DN T cells represented the majority of T cells. Flow cytometric analysis of single-cell suspensions from fresh tumors identified DN T cells as predominantly Vδ2- γδ T cells. In the context of γδ T-cell inflammation, these cells expressed PD-1 and LAG3, which is consistent with a suppressed or exhausted phenotype, and CD103, which indicates tissue residency. Furthermore, single-cell RNA sequencing (scRNA-seq) identified a transcriptional profile of γδ T cells suggestive of proinflammatory potential. T-cell receptor (TCR) analysis confirmed clonal expansion of Vδ1 and Vδ3 clonotypes, and functional studies using cloned γδ TCRs demonstrated restriction of these for CD1c and MR1 antigen-presenting molecules. On the basis of a 13-gene γδ T-cell signature derived from scRNA-seq analysis, gene-set enrichment on bulk RNA-seq data showed a positive correlation between enrichment scores and DN T-cell infiltrates. An improved disease-specific survival was evident for patients with high enrichment scores, and complete responses to anti-PD-1/PD-L1 treatment were observed in three of four cases with high enrichment scores. Thus, γδ T-cell infiltration may serve as a prognostic biomarker and should be explored for therapeutic interventions.See related Spotlight on p. 600.

Towards Routine Implementation of Liquid Biopsies in Cancer Management: It Is Always Too Early, until Suddenly It Is Too Late.

Blood-based liquid biopsies are considered a new and promising diagnostic and monitoring tool for cancer. As liquid biopsies only require a blood draw, they are non-invasive, potentially more rapid and assumed to be a less costly alternative to genomic analysis of tissue biopsies. A multi-disciplinary workshop (n = 98 registrations) was organized to discuss routine implementation of liquid biopsies in cancer management. Real-time polls were used to engage with experts' about the current evidence of clinical utility and the barriers to implementation of liquid biopsies. Clinical, laboratory and health economics presentations were given to illustrate the opportunities and current levels of evidence, followed by three moderated break-out sessions to discuss applications. The workshop concluded that tumor-informed assays using next-generation sequencing (NGS) or PCR-based genotyping assays will most likely provide better clinical utility than tumor-agnostic assays, yet at a higher cost. For routine application, it will be essential to determine clinical utility, to define the minimum quality standards and performance of testing platforms and to ensure their use is integrated into current clinical workflows including how they complement tissue biopsies and imaging. Early health economic models may help identifying the most viable application of liquid biopsies. Alternative funding models for the translation of complex molecular diagnostics, such as liquid biopsies, may also be explored if clinical utility has been demonstrated and when their use is recommended in multi-disciplinary consensus guidelines.

FAPI PET/CT: Will It End the Hegemony of 18F-FDG in Oncology?

For over 40 years, 18F-FDG has been the dominant PET tracer in neurology, cardiology, inflammatory diseases, and, most particularly, oncology. Combined with the ability to perform whole-body scanning, 18F-FDG has revolutionized the evaluation of cancer and has stifled the adoption of other tracers, except in situations where low avidity or high background activity limits diagnostic performance. The strength of 18F-FDG has generally been its ability to detect disease in the absence of structural abnormality, thereby enhancing diagnostic sensitivity, but its simultaneous weakness has been a lack of specificity due to diverse pathologies with enhanced glycolysis. Radiotracers that leverage other hallmarks of cancer or specific cell-surface targets are gradually finding a niche in the diagnostic armamentarium. However, none have had sufficient sensitivity to realistically compete with 18F-FDG for evaluation of the broad spectrum of malignancies. Perhaps, this situation is about to change with development of a class of tracers targeting fibroblast activation protein that have low uptake in almost all normal tissues but high uptake in most cancer types. In this review, the development and exciting preliminary clinical data relating to various fibroblast activation protein-specific small-molecule inhibitor tracers in oncology will be discussed along with potential nononcologic applications.

Clinical, FDG-PET and molecular markers of immune checkpoint inhibitor response in patients with metastatic Merkel cell carcinoma.

BACKGROUND: Metastatic Merkel cell carcinoma (mMCC) is an aggressive neuroendocrine malignancy of the skin with a poor prognosis. Immune checkpoint inhibitors (ICIs) have shown substantial efficacy and favorable safety in clinical trials. METHODS: Medical records of patients (pts) with mMCC treated with ICIs from August 2015 to December 2018 at Peter MacCallum Cancer Centre in Australia were analyzed. Response was assessed with serial imaging, the majority with FDG-PET/CT scans. RNA sequencing and immunohistochemistry for PD-L1, CD3 and Merkel cell polyomavirus (MCPyV) on tumor samples was performed. RESULTS: 23 pts with mMCC were treated with ICIs. A median of 8 cycles (range 1 to 47) were administered, with treatment ongoing in 6 pts. Objective responses (OR) were observed in 14 pts (61%): 10 (44%) complete responses (CR) and 4 (17%) partial responses (PR). Median time to response was 8 weeks (range 6 to 12) and 12-month progression-free survival rate was 39%. Increased OR were seen in pts aged less than 75 (OR 80% vs 46%), no prior history of chemotherapy (OR 64% vs 50%), patients with an immune-related adverse event (OR 100% vs 43%) and in MCPyV-negative tumors (OR 69% vs 43%). Pts with a CR had lower mean metabolic tumor volume on baseline FDG-PET/CT scan (CR: 35.7 mL, no CR: 187.8 mL, p=0.05). There was no correlation between PD-L1 positivity and MCPyV status (p=0.764) or OR (p=0.245). 10 pts received radiation therapy (RT) during ICI: 4 pts started RT concurrently (OR 75%, CR 50%), 3 pts had isolated ICI-resistant lesions successfully treated with RT and 3 pts with multisite progression continued to progress despite RT. Overall, 6 pts (26%) had grade 1-2 immune-related adverse events. CONCLUSION: ICIs showed efficacy and safety in mMCC consistent with trial data. Clinical and imaging predictors of response were identified.

CUP-AI-Dx: A tool for inferring cancer tissue of origin and molecular subtype using RNA gene-expression data and artificial intelligence.

BACKGROUND: Cancer of unknown primary (CUP), representing approximately 3-5% of all malignancies, is defined as metastatic cancer where a primary site of origin cannot be found despite a standard diagnostic workup. Because knowledge of a patient's primary cancer remains fundamental to their treatment, CUP patients are significantly disadvantaged and most have a poor survival outcome. Developing robust and accessible diagnostic methods for resolving cancer tissue of origin, therefore, has significant value for CUP patients. METHODS: We developed an RNA-based classifier called CUP-AI-Dx that utilizes a 1D Inception convolutional neural network (1D-Inception) model to infer a tumor's primary tissue of origin. CUP-AI-Dx was trained using the transcriptional profiles of 18,217 primary tumours representing 32 cancer types from The Cancer Genome Atlas project (TCGA) and International Cancer Genome Consortium (ICGC). Gene expression data was ordered by gene chromosomal coordinates as input to the 1D-CNN model, and the model utilizes multiple convolutional kernels with different configurations simultaneously to improve generality. The model was optimized through extensive hyperparameter tuning, including different max-pooling layers and dropout settings. For 11 tumour types, we also developed a random forest model that can classify the tumour's molecular subtype according to prior TCGA studies. The optimised CUP-AI-Dx tissue of origin classifier was tested on 394 metastatic samples from 11 tumour types from TCGA and 92 formalin-fixed paraffin-embedded (FFPE) samples representing 18 cancer types from two clinical laboratories. The CUP-AI-Dx molecular subtype was also independently tested on independent ovarian and breast cancer microarray datasets FINDINGS: CUP-AI-Dx identifies the primary site with an overall top-1-accuracy of 98.54% in cross-validation and 96.70% on a test dataset. When applied to two independent clinical-grade RNA-seq datasets generated from two different institutes from the US and Australia, our model predicted the primary site with a top-1-accuracy of 86.96% and 72.46% respectively. INTERPRETATION: The CUP-AI-Dx predicts tumour primary site and molecular subtype with high accuracy and therefore can be used to assist the diagnostic work-up of cancers of unknown primary or uncertain origin using a common and accessible genomics platform. FUNDING: NIH R35 GM133562, NCI P30 CA034196, Victorian Cancer Agency Australia.