GAiN: An integrative tool utilizing generative adversarial neural networks for augmented gene expression analysis.

Big genomic data and artificial intelligence (AI) are ushering in an era of precision medicine, providing opportunities to study previously under-represented subtypes and rare diseases rather than categorize them as variances. However, clinical researchers face challenges in accessing such novel technologies as well as reliable methods to study small datasets or subcohorts with unique phenotypes. To address this need, we developed an integrative approach, GAiN, to capture patterns of gene expression from small datasets on the basis of an ensemble of generative adversarial networks (GANs) while leveraging big population data. Where conventional biostatistical methods fail, GAiN reliably discovers differentially expressed genes (DEGs) and enriched pathways between two cohorts with limited numbers of samples (n = 10) when benchmarked against a gold standard. GAiN is freely available at GitHub. Thus, GAiN may serve as a crucial tool for gene expression analysis in scenarios with limited samples, as in the context of rare diseases, under-represented populations, or limited investigator resources.

Early detection of malignant and pre-malignant peripheral nerve tumors using cell-free DNA fragmentomics.

Early detection of neurofibromatosis type 1 (NF1) associated peripheral nerve sheath tumors (PNST) informs clinical decision-making, potentially averting deadly outcomes. Here, we describe a cell-free DNA (cfDNA) fragmentomic approach which distinguishes non-malignant, pre-malignant and malignant forms of NF1 PNST. Using plasma samples from a novel cohort of 101 NF1 patients and 21 healthy controls, we validated that our previous cfDNA copy number alteration (CNA)-based approach identifies malignant peripheral nerve sheath tumor (MPNST) but cannot distinguish among benign and premalignant states. We therefore investigated the ability of fragment-based cfDNA features to differentiate NF1-associated tumors including binned genome-wide fragment length ratios, end motif analysis, and non-negative matrix factorization deconvolution of fragment lengths. Fragmentomic methods were able to differentiate pre-malignant states including atypical neurofibromas (AN). Fragmentomics also adjudicated AN cases suspicious for MPNST, correctly diagnosing samples noninvasively, which could have informed clinical management. Overall, this study pioneers the early detection of malignant and premalignant peripheral nerve sheath tumors in NF1 patients using plasma cfDNA fragmentomics. In addition to screening applications, this novel approach distinguishes atypical neurofibromas from benign plexiform neurofibromas and malignant peripheral nerve sheath tumors, enabling more precise clinical diagnosis and management.

Sensitive MRD Detection from Lymphatic Fluid after Surgery in HPV-Associated Oropharyngeal Cancer.

PURPOSE: Our goal was to demonstrate that lymphatic drainage fluid (lymph) has improved sensitivity in quantifying postoperative minimal residual disease (MRD) in locally advanced human papillomavirus (HPV)-associated oropharyngeal squamous cell carcinoma (OPSCC) compared with plasma, and leverage this novel biofluid for patient risk stratification. EXPERIMENTAL DESIGN: We prospectively collected lymph samples from neck drains of 106 patients with HPV (+) OPSCC, along with 67 matched plasma samples, 24 hours after surgery. PCR and next-generation sequencing were used to quantify cancer-associated cell-free HPV (cf-HPV) and tumor-informed variants in lymph and plasma. Next, lymph cf-HPV and variants were compared with TNM stage, extranodal extension (ENE), and composite definitions of high-risk pathology. We then created a machine learning model, informed by lymph MRD and clinicopathologic features, to compare with progression-free survival (PFS). RESULTS: Postoperative lymph was enriched with cf-HPV compared with plasma (P < 0.0001) and correlated with pN2 stage (P = 0.003), ENE (P < 0.0001), and trial-defined pathologic risk criteria (mean AUC = 0.78). In addition, the lymph mutation number and variant allele frequency were higher in pN2 ENE (+) necks than in pN1 ENE (+) (P = 0.03, P = 0.02) or pN0-N1 ENE (-) (P = 0.04, P = 0.03, respectively). The lymph MRD-informed risk model demonstrated inferior PFS in high-risk patients (AUC = 0.96, P < 0.0001). CONCLUSIONS: Variant and cf-HPV quantification, performed in 24-hour postoperative lymph samples, reflects single- and multifeature high-risk pathologic criteria. Incorporating lymphatic MRD and clinicopathologic feature analysis can stratify PFS early after surgery in patients with HPV (+) head and neck cancer. See related commentary by Shannon and Iyer, p. 1223.

Genomic approaches to cancer and minimal residual disease detection using circulating tumor DNA.

Liquid biopsies using cell-free circulating tumor DNA (ctDNA) are being used frequently in both research and clinical settings. ctDNA can be used to identify actionable mutations to personalize systemic therapy, detect post-treatment minimal residual disease (MRD), and predict responses to immunotherapy. ctDNA can also be isolated from a range of different biofluids, with the possibility of detecting locoregional MRD with increased sensitivity if sampling more proximally than blood plasma. However, ctDNA detection remains challenging in early-stage and post-treatment MRD settings where ctDNA levels are minuscule giving a high risk for false negative results, which is balanced with the risk of false positive results from clonal hematopoiesis. To address these challenges, researchers have developed ever-more elegant approaches to lower the limit of detection (LOD) of ctDNA assays toward the part-per-million range and boost assay sensitivity and specificity by reducing sources of low-level technical and biological noise, and by harnessing specific genomic and epigenomic features of ctDNA. In this review, we highlight a range of modern assays for ctDNA analysis, including advancements made to improve the signal-to-noise ratio. We further highlight the challenge of detecting ultra-rare tumor-associated variants, overcoming which will improve the sensitivity of post-treatment MRD detection and open a new frontier of personalized adjuvant treatment decision-making.

Genomic Patterns of Malignant Peripheral Nerve Sheath Tumor (MPNST) Evolution Correlate with Clinical Outcome and Are Detectable in Cell-Free DNA.

Malignant peripheral nerve sheath tumor (MPNST), an aggressive soft-tissue sarcoma, occurs in people with neurofibromatosis type 1 (NF1) and sporadically. Whole-genome and multiregional exome sequencing, transcriptomic, and methylation profiling of 95 tumor samples revealed the order of genomic events in tumor evolution. Following biallelic inactivation of NF1, loss of CDKN2A or TP53 with or without inactivation of polycomb repressive complex 2 (PRC2) leads to extensive somatic copy-number aberrations (SCNA). Distinct pathways of tumor evolution are associated with inactivation of PRC2 genes and H3K27 trimethylation (H3K27me3) status. Tumors with H3K27me3 loss evolve through extensive chromosomal losses followed by whole-genome doubling and chromosome 8 amplification, and show lower levels of immune cell infiltration. Retention of H3K27me3 leads to extensive genomic instability, but an immune cell-rich phenotype. Specific SCNAs detected in both tumor samples and cell-free DNA (cfDNA) act as a surrogate for H3K27me3 loss and immune infiltration, and predict prognosis. SIGNIFICANCE: MPNST is the most common cause of death and morbidity for individuals with NF1, a relatively common tumor predisposition syndrome. Our results suggest that somatic copy-number and methylation profiling of tumor or cfDNA could serve as a biomarker for early diagnosis and to stratify patients into prognostic and treatment-related subgroups. This article is highlighted in the In This Issue feature, p. 517.

Correction: Urine tumor DNA detection of minimal residual disease in muscle-invasive bladder cancer treated with curative-intent radical cystectomy: A cohort study.

[This corrects the article DOI: 10.1371/journal.pmed.1003732.].

Cell-free DNA ultra-low-pass whole genome sequencing to distinguish malignant peripheral nerve sheath tumor (MPNST) from its benign precursor lesion: A cross-sectional study.

BACKGROUND: The leading cause of mortality for patients with the neurofibromatosis type 1 (NF1) cancer predisposition syndrome is the development of malignant peripheral nerve sheath tumor (MPNST), an aggressive soft tissue sarcoma. In the setting of NF1, this cancer type frequently arises from within its common and benign precursor, plexiform neurofibroma (PN). Transformation from PN to MPNST is challenging to diagnose due to difficulties in distinguishing cross-sectional imaging results and intralesional heterogeneity resulting in biopsy sampling errors. METHODS AND FINDINGS: This multi-institutional study from the National Cancer Institute and Washington University in St. Louis used fragment size analysis and ultra-low-pass whole genome sequencing (ULP-WGS) of plasma cell-free DNA (cfDNA) to distinguish between MPNST and PN in patients with NF1. Following in silico enrichment for short cfDNA fragments and copy number analysis to estimate the fraction of plasma cfDNA originating from tumor (tumor fraction), we developed a noninvasive classifier that differentiates MPNST from PN with 86% pretreatment accuracy (91% specificity, 75% sensitivity) and 89% accuracy on serial analysis (91% specificity, 83% sensitivity). Healthy controls without NF1 (participants = 16, plasma samples = 16), PN (participants = 23, plasma samples = 23), and MPNST (participants = 14, plasma samples = 46) cohorts showed significant differences in tumor fraction in plasma (P = 0.001) as well as cfDNA fragment length (P < 0.001) with MPNST samples harboring shorter fragments and being enriched for tumor-derived cfDNA relative to PN and healthy controls. No other covariates were significant on multivariate logistic regression. Mutational analysis demonstrated focal NF1 copy number loss in PN and MPNST patient plasma but not in healthy controls. Greater genomic instability including alterations associated with malignant transformation (focal copy number gains in chromosome arms 1q, 7p, 8q, 9q, and 17q; focal copy number losses in SUZ12, SMARCA2, CDKN2A/B, and chromosome arms 6p and 9p) was more prominently observed in MPNST plasma. Furthermore, the sum of longest tumor diameters (SLD) visualized by cross-sectional imaging correlated significantly with paired tumor fractions in plasma from MPNST patients (r = 0.39, P = 0.024). On serial analysis, tumor fraction levels in plasma dynamically correlated with treatment response to therapy and minimal residual disease (MRD) detection before relapse. Study limitations include a modest MPNST sample size despite accrual from 2 major referral centers for this rare malignancy, and lack of uniform treatment and imaging protocols representing a real-world cohort. CONCLUSIONS: Tumor fraction levels derived from cfDNA fragment size and copy number alteration analysis of plasma cfDNA using ULP-WGS significantly correlated with MPNST tumor burden, accurately distinguished MPNST from its benign PN precursor, and dynamically correlated with treatment response. In the future, our findings could form the basis for improved early cancer detection and monitoring in high-risk cancer-predisposed populations.

Urine tumor DNA detection of minimal residual disease in muscle-invasive bladder cancer treated with curative-intent radical cystectomy: A cohort study.

BACKGROUND: The standard of care treatment for muscle-invasive bladder cancer (MIBC) is radical cystectomy, which is typically preceded by neoadjuvant chemotherapy. However, the inability to assess minimal residual disease (MRD) noninvasively limits our ability to offer bladder-sparing treatment. Here, we sought to develop a liquid biopsy solution via urine tumor DNA (utDNA) analysis. METHODS AND FINDINGS: We applied urine Cancer Personalized Profiling by Deep Sequencing (uCAPP-Seq), a targeted next-generation sequencing (NGS) method for detecting utDNA, to urine cell-free DNA (cfDNA) samples acquired between April 2019 and November 2020 on the day of curative-intent radical cystectomy from 42 patients with localized bladder cancer. The average age of patients was 69 years (range: 50 to 86), of whom 76% (32/42) were male, 64% (27/42) were smokers, and 76% (32/42) had a confirmed diagnosis of MIBC. Among MIBC patients, 59% (19/32) received neoadjuvant chemotherapy. utDNA variant calling was performed noninvasively without prior sequencing of tumor tissue. The overall utDNA level for each patient was represented by the non-silent mutation with the highest variant allele fraction after removing germline variants. Urine was similarly analyzed from 15 healthy adults. utDNA analysis revealed a median utDNA level of 0% in healthy adults and 2.4% in bladder cancer patients. When patients were classified as those who had residual disease detected in their surgical sample (n = 16) compared to those who achieved a pathologic complete response (pCR; n = 26), median utDNA levels were 4.3% vs. 0%, respectively (p = 0.002). Using an optimal utDNA threshold to define MRD detection, positive utDNA MRD detection was highly correlated with the absence of pCR (p < 0.001) with a sensitivity of 81% and specificity of 81%. Leave-one-out cross-validation applied to the prediction of pathologic response based on utDNA MRD detection in our cohort yielded a highly significant accuracy of 81% (p = 0.007). Moreover, utDNA MRD-positive patients exhibited significantly worse progression-free survival (PFS; HR = 7.4; 95% CI: 1.4-38.9; p = 0.02) compared to utDNA MRD-negative patients. Concordance between urine- and tumor-derived mutations, determined in 5 MIBC patients, was 85%. Tumor mutational burden (TMB) in utDNA MRD-positive patients was inferred from the number of non-silent mutations detected in urine cfDNA by applying a linear relationship derived from The Cancer Genome Atlas (TCGA) whole exome sequencing of 409 MIBC tumors. We suggest that about 58% of these patients with high inferred TMB might have been candidates for treatment with early immune checkpoint blockade. Study limitations included an analysis restricted only to single-nucleotide variants (SNVs), survival differences diminished by surgery, and a low number of DNA damage response (DRR) mutations detected after neoadjuvant chemotherapy at the MRD time point. CONCLUSIONS: utDNA MRD detection prior to curative-intent radical cystectomy for bladder cancer correlated significantly with pathologic response, which may help select patients for bladder-sparing treatment. utDNA MRD detection also correlated significantly with PFS. Furthermore, utDNA can be used to noninvasively infer TMB, which could facilitate personalized immunotherapy for bladder cancer in the future.

ctDNA MRD Detection and Personalized Oncogenomic Analysis in Oligometastatic Colorectal Cancer From Plasma and Urine.

We hypothesized that circulating tumor DNA (ctDNA) molecular residual disease (MRD) analysis without prior mutational knowledge could be performed after neoadjuvant chemotherapy to assess oligometastatic colorectal cancer (CRC) treated surgically with curative intent. We also investigated urine as an alternative analyte for ctDNA MRD detection in this nongenitourinary setting. PATIENTS AND METHODS: We applied AVENIO targeted next-generation sequencing to plasma, tumor, and urine samples acquired on the day of curative-intent surgery from 24 prospectively enrolled patients with oligometastatic CRC. Age-related clonal hematopoiesis was accounted for by removing variants also present in white blood cells. Plasma and urine ctDNA MRD were correlated with tumor cells detected in the surgical specimen, and adjuvant treatment strategies were proposed based on ctDNA-inferred tumor mutational burden (iTMB) and targetable alterations. RESULTS: Seventy-one percent of patients were treated with neoadjuvant chemotherapy. Tumor-naive plasma ctDNA analysis detected MRD at a median level of 0.62% with 95% sensitivity and 100% specificity, and 94% and 77% sensitivity when only considering patients treated with neoadjuvant chemotherapy and putative driver mutations, respectively. In urine, ctDNA MRD detection specificity remained high at 100%, but sensitivity decreased to 64% with median levels being 11-fold lower than in plasma (P < .0001). Personalized ctDNA MRD oncogenomic analysis revealed 81% of patients might have been candidates for adjuvant immunotherapy based on high iTMB or targeted therapy based on actionable PIK3CA mutations. CONCLUSION: Tumor-naive plasma ctDNA analysis can sensitively and specifically detect MRD in patients with oligometastatic CRC after neoadjuvant chemotherapy. Urine-based ctDNA MRD detection is also feasible; however, it is less sensitive than plasma because of significantly lower levels. Oligometastatic patients with detectable MRD may benefit from additional personalized treatment based on ctDNA-derived oncogenomic profiling.

Once-Per-Visit Alerts: A Means to Study Alert Compliance and Reduce Repeat Laboratory Testing.

BACKGROUND: Clinical decision support alerts for laboratory testing have poor compliance. Once-per-visit alerts, triggered by reorder of a test within the same admission, are highly specific for unnecessary orders and provide a means to study alert compliance. METHODS: Once-per-visit alerts for 18 laboratory orderables were analyzed over a 60-month period from September 2012 to October 2016 at a 1200-bed academic medical center. To determine correlates of alert compliance, we compared alerts by test and provider characteristics. RESULTS: Overall alert compliance was 54.5%. In multivariate regression, compliance correlated with length of stay at time of alert, provider type, previous alerts in a patient visit, test ordered, total alerts experienced by ordering provider, and previous order status. CONCLUSIONS: A diverse set of provider and test characteristics influences compliance with once-per-visit laboratory alerts. Future alerts should incorporate these characteristics into alert design to minimize alert overrides.

Short- and Long-term Biologic Variability of Galectin-3 and Other Cardiac Biomarkers in Patients with Stable Heart Failure and Healthy Adults.

BACKGROUND: Galectin-3 (Gal-3) has been suggested as a prognostic biomarker in heart failure (HF) patients that may better reflect disease progression than traditional markers, including B-type natriuretic peptide (BNP) and cardiac troponins. To fully establish the utility of any biomarker in HF, its biologic variability must be characterized. METHODS: To assess biologic variability, 59 patients were prospectively recruited, including 23 male and 16 female patients with stable HF and 10 male and 10 female healthy individuals. Gal-3, BNP, and high-sensitivity cardiac troponin I (hs-cTnI) were assayed at 5 time points within a 3-week period to assess short-term biologic variability. Long-term (3-month) biologic variability was assessed with samples collected at enrollment and after 4, 8, and 12 weeks. RESULTS: Among healthy individuals, mean short-term biologic variability, expressed as intraindividual CV (CVI), was 4.5% for Gal-3, 29.0% for BNP, and 14.5% for hs-cTnI; long-term biologic variability was 5.5% for Gal-3, 34.7% for BNP, and 14.7% for hs-cTnI. In stable HF patients, mean short-term biologic variability was 7.1% for Gal-3, 22.5% for BNP, and 8.5% for hs-cTnI, and mean long-term biologic variability was 7.7% for Gal-3, 27.6% for BNP, and 9.6% for hs-cTnI. CONCLUSIONS: The finding that Gal-3 has minimal intraindividual biological variability adds to its potential as a useful biomarker in HF patients.

Incidence of humoral hypercalcemia of malignancy among hypercalcemic patients with cancer.

BACKGROUND: Malignancy-associated hypercalcemia (MAHC) is the most common cause of hypercalcemia among hospitalized patients. MAHC can result from the production of parathyroid hormone related peptide (PTHrP) which is known as humoral hypercalcemia of malignancy (HHM). HHM is commonly thought to account for approximately 80% of MAHC. METHODS: We conducted a 12-year review of PTHrP testing at our institution to establish the prevalence of HHM among patients with MAHC. RESULTS: A total of 524 PTHrP immunoassays were performed during the study period of which 470 tests qualified for inclusion in the analysis. Evidence of malignancy was found for 242 of 470 patients (51%). No etiology could be determined for 98 cases of MAHC (40%) and increased PTHrP contributed to 92 cases (38%) of MAHC. Age, race and gender were not associated with HHM. Increased PTHrP was observed at initial malignancy diagnosis in 20% of cases. PTHrP was never increased outside of the context of malignancy. DISCUSSION: The prevalence of HHM among patients with MAHC is likely to be lower than previously described.

Impact on Patient Management and Outcome of Switching between 2 Contemporary Sensitive Cardiac Troponin Assays.

BACKGROUND: Myocardial infarction is characterized by an increase of cardiac troponin I (cTnI) above the 99th percentile of a reference population. Our hospital switched from 1 contemporary cTnI assay to another and observed a doubling of cTnI results above the assays' respective 99th percentile cutoffs. We investigated the potential impact on inpatient management and outcomes. METHODS: We performed a retrospective cohort study of 45 498 individuals with ≥1 cTnI result between January 2013 and June 2014. The Dimension cTnI assay was used in 2013; the Abbott Architect cTnI assay was used in 2014. RESULTS: Before switching cTnI assays, 19.2% (4742/30 872) of patients had at least 1 of the first 3 cTnIs above the 99th percentile (0.07 µg/L). After switching to the Architect cTnI assay, 31.4% (4034/14 626) of patients had at least 1 cTnI above the 99th percentile (0.03 µg/L). This increase was due to the difference in the assays' 99th percentile cutoffs. Having an increased cTnI reported on the Architect assay that would not have been reported as such on the Dimension assay (0.03-0.06 µg/L) correlated with increased inpatient mortality, length of stay, non-ST elevation myocardial infarction diagnosis, therapeutic heparin use, and percutaneous coronary intervention, relative to individuals with cTnI <0.03 µg/L. CONCLUSIONS: The changes observed in patient outcomes and management were likely due to the increased sensitivity and lower 99th percentile cutoff of the Architect assay. It is important to recognize the potential impact that differences in sensitivity and assay configuration may have on patient management.

HuR function and translational state analysis following global brain ischemia and reperfusion.

Prolonged translation arrest in post-ischemic hippocampal CA1 pyramidal neurons precludes translation of induced stress genes and directly correlates with cell death. We evaluated the regulation of mRNAs containing adenine- and uridine-rich elements (ARE) by assessing HuR protein and hsp70 mRNA nuclear translocation, HuR polysome binding, and translation state analysis of CA1 and CA3 at 8 h of reperfusion after 10 min of global cerebral ischemia. There was no difference between CA1 and CA3 at 8 h of reperfusion in nuclear or cytoplasmic HuR protein or hsp70 mRNA, or HuR polysome association, suggesting that neither mechanism contributed to post-ischemic outcome. Translation state analysis revealed that 28 and 58 % of unique mRNAs significantly different between 8hR and NIC, in CA3 and CA1, respectively, were not polysome-bound. There was significantly greater diversity of polysome-bound mRNAs in reperfused CA3 compared to CA1, and in both regions, ARE-containing mRNAs accounted for 4-5 % of the total. These data indicate that posttranscriptional ARE-containing mRNA regulation occurs in reperfused neurons and contributes to post-ischemic outcome. Understanding the differential responses of vulnerable and resistant neurons to ischemia will contribute to the development of effective neuroprotective therapies.

Texture analysis of poly-adenylated mRNA staining following global brain ischemia and reperfusion.

Texture analysis provides a means to quantify complex changes in microscope images. We previously showed that cytoplasmic poly-adenylated mRNAs form mRNA granules in post-ischemic neurons and that these granules correlated with protein synthesis inhibition and hence cell death. Here we utilized the texture analysis software MaZda to quantify mRNA granules in photomicrographs of the pyramidal cell layer of rat hippocampal region CA3 around 1h of reperfusion after 10min of normothermic global cerebral ischemia. At 1h reperfusion, we observed variations in the texture of mRNA granules amongst samples that were readily quantified by texture analysis. Individual sample variation was consistent with the interpretation that animal-to-animal variations in mRNA granules reflected the time-course of mRNA granule formation. We also used texture analysis to quantify the effect of cycloheximide, given either before or after brain ischemia, on mRNA granules. If administered before ischemia, cycloheximide inhibited mRNA granule formation, but if administered after ischemia did not prevent mRNA granulation, indicating mRNA granule formation is dependent on dissociation of polysomes. We conclude that texture analysis is an effective means for quantifying the complex morphological changes induced in neurons by brain ischemia and reperfusion.

Translation arrest and ribonomics in post-ischemic brain: layers and layers of players.

A persistent translation arrest (TA) correlates precisely with the selective vulnerability of post-ischemic neurons. Mechanisms of post-ischemic TA that have been assessed include ribosome biochemistry, the link between TA and stress responses, and the inactivation of translational components via sequestration in subcellular structures. Each of these approaches provides a perspective on post-ischemic TA. Here, we develop the notion that mRNA regulation via RNA-binding proteins, or ribonomics, also contributes to post-ischemic TA. We describe the ribonomic network, or structures involved in mRNA regulation, including nuclear foci, polysomes, stress granules, embryonic lethal abnormal vision/Hu granules, processing bodies, exosomes, and RNA granules. Transcriptional, ribonomic, and ribosomal regulation together provide multiple layers mediating cell reprogramming. Stress gene induction via the heat-shock response, immediate early genes, and endoplasmic reticulum stress represents significant reprogramming of post-ischemic neurons. We present a model of post-ischemic TA in ischemia-resistant neurons that incorporates ribonomic considerations. In this model, selective translation of stress-induced mRNAs contributes to translation recovery. This model provides a basis to study dysfunctional stress responses in vulnerable neurons, with a key focus on the inability of vulnerable neurons to selectively translate stress-induced mRNAs. We suggest a ribonomic approach will shed new light on the roles of mRNA regulation in persistent TA in vulnerable post-ischemic neurons.