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 data set of 39,787 solid tumors sequenced using a clinically 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, rivaling the performance of WGS-based methods. GDD-ENS can also guide diagnoses of 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. SIGNIFICANCE: We describe a highly accurate tumor-type prediction model, designed specifically for clinical implementation. Our model relies only on widely used cancer gene panel sequencing data, predicts across 38 distinct cancer types, and supports integration of patient-specific nongenomic information for enhanced decision support in challenging diagnostic situations. See related commentary by Garg, p. 906. This article is featured in Selected Articles from This Issue, p. 897.

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.

Cigarette Smoking and Opium Use in Relation to the Oral Microbiota in Iran.

Cigarettes and opium contain chemicals and particulate matter that may modify the oral microbiota. This study aimed to investigate the association between cigarette and opium use with the oral microbiota. A total of 558 participants were recruited from Iran between 2011 and 2015. Individuals were categorized as never cigarette nor opium users, ever cigarette-only smokers, ever opium-only users, and ever both cigarette and opium users. Participants provided saliva samples for 16S rRNA gene sequencing. Logistic regression, microbiome regression-based kernel association test (MiRKAT), and zero-inflated beta regression models were calculated. For every increase in 10 observed amplicon sequence variants (ASVs), the odds for being a cigarette-only smoker, opium-only user, and both user compared to never users decreased by 9% (odds ratio [OR] = 0.91; 95% confidence interval [95% CI] = 0.86 to 0.97), 13% (OR = 0.87; 95% CI = 0.75 to 1.01), and 12% (OR = 0.88; 95% CI = 0.80 to 0.96), respectively. The microbial communities differed by cigarette and opium use as indicated by MiRKAT models testing the three beta-diversity matrices (P < 0.05 for all). Three genera were less likely and one genus was more likely to be detected in cigarette-only smokers or opium-only users than in never users. The relative abundance of the phylum Actinobacteria (never, 14.78%; both, 21.20%) was higher and the phyla Bacteroidetes (never, 17.63%; both, 11.62%) and Proteobacteria (never, 9.06%; both, 3.70%) were lower in users of both cigarettes and opium, while the phylum Firmicutes (never, 54.29%; opium, 65.49%) was higher in opium-only users. Cigarette and opium use was associated with lower alpha-diversity, overall oral microbiota community composition, and both the presence and relative abundance of multiple taxa. IMPORTANCE Cigarette smoking and opium use are associated with periodontal disease caused by specific bacteria such as Porphyromonas gingivalis, which suggests a link between cigarette smoking and opium use and the oral microbiota. Alterations of the oral microbiota in cigarette smokers compared to nonsmokers have been reported, but this has not been studied across diverse populations. Additionally, the association of opium use with the oral microbiota has not been investigated to date. We conducted this study to investigate differences in the oral microbiota between ever users of cigarettes only, opium only, and both cigarettes and opium and never users of cigarettes and opium in Iran. Lower alpha-diversity, distinct overall oral microbial communities, and the presence and relative abundance of multiple taxa have been found for users of cigarettes and/or opium.

Lack of transgenerational effects of ionizing radiation exposure from the Chernobyl accident.

Effects of radiation exposure from the Chernobyl nuclear accident remain a topic of interest. We investigated germline de novo mutations (DNMs) in children born to parents employed as cleanup workers or exposed to occupational and environmental ionizing radiation after the accident. Whole-genome sequencing of 130 children (born 1987-2002) and their parents did not reveal an increase in the rates, distributions, or types of DNMs relative to the results of previous studies. We find no elevation in total DNMs, regardless of cumulative preconception gonadal paternal [mean = 365 milligrays (mGy), range = 0 to 4080 mGy] or maternal (mean = 19 mGy, range = 0 to 550 mGy) exposure to ionizing radiation. Thus, we conclude that, over this exposure range, evidence is lacking for a substantial effect on germline DNMs in humans, suggesting minimal impact from transgenerational genetic effects.

Comparison of Oral Microbiota Collected Using Multiple Methods and Recommendations for New Epidemiologic Studies.

Epidemiologic studies use various biosample collection methods to study associations between human oral microbiota and health outcomes. However, the agreement between the different methods is unclear. We compared a commercially available OMNIgene ORAL kit to three alternative collection methods: Saccomanno's fixative, Scope mouthwash, and nonethanol mouthwash. Oral samples were collected from 40 individuals over 4 visits. Two samples were collected from each subject per visit: one with OMNIgene and one with an alternative method. DNA was extracted using the DSP DNA Virus Pathogen kit, and the V4 region of the 16S rRNA gene was PCR amplified and sequenced using MiSeq. Oral collection methods were compared based on alpha and beta diversity metrics and phylum- and genus-level relative abundances. All alpha diversity metrics were significantly lower for Saccomanno's fixative than for OMNIgene (P < 0.001), whereas the two mouthwashes were more similar to OMNIgene. Principal-coordinate analysis (PCoA) using the Bray-Curtis and weighted UniFrac beta diversity matrices showed large differences in the microbial compositions of samples collected with Saccomanno's compared to those with OMNIgene and the mouthwashes. Clustering by collection method was not observed in unweighted UniFrac PCoA plots, suggesting differences in relative abundances but not specific taxa detected by the collection methods. Relative abundances of most taxa were significantly different between OMNIgene and the other methods at each taxonomic level, with Saccomanno's showing the least agreement with OMNIgene. There were clear differences in oral microbial communities between the four oral collection methods, particularly for Saccomanno's fixative.IMPORTANCE We compared four different oral collection methods for studying the human oral microbiome: an OMNIgene ORAL kit, Scope mouthwash, nonethanol mouthwash, and Saccomanno's fixative. Our study shows that the type of the collection method can have a large impact on the results of an oral microbiome analysis. We recommend that one consistent oral collection method should be used for all oral microbiome comparisons. While Scope and nonethanol mouthwashes are less expensive and provide results similar to those with OMNIgene, Saccomanno's fixative may be unfavorable due to the microbial differences detected in this study. Our results will help guide the design of future oral microbiome studies.

Oral microbial community composition is associated with pancreatic cancer: A case-control study in Iran.

BACKGROUND: Oral microbiota may be related to pancreatic cancer risk because periodontal disease, a condition linked to multiple specific microbes, has been associated with increased risk of pancreatic cancer. We evaluated the association between oral microbiota and pancreatic cancer in Iran. METHODS: A total of 273 pancreatic adenocarcinoma cases and 285 controls recruited from tertiary hospitals and a specialty clinic in Tehran, Iran provided saliva samples and filled out a questionnaire regarding demographics and lifestyle characteristics. DNA was extracted from saliva and the V4 region of the 16S rRNA gene was PCR amplified and sequenced on the MiSeq. The sequencing data were processed using the DADA2 plugin in QIIME 2 and taxonomy was assigned against the Human Oral Microbiome Database. Logistic regression and MiRKAT models were calculated with adjustment for potential confounders. RESULTS: No association was observed for alpha diversity with an average of 91.11 (standard deviation [SD] 2.59) sequence variants for cases and 89.42 (SD 2.58) for controls. However, there was evidence for an association between beta diversity and case status. The association between the Bray-Curtis dissimilarity and pancreatic cancer was particularly strong with a MiRKAT P-value of .000142 and specific principal coordinate vectors had strong associations with cancer risk. Several specific taxa were also associated with case status after adjustment for multiple comparisons. CONCLUSION: The overall microbial community appeared to differ between pancreatic cancer cases and controls. Whether these reflect differences evident before development of pancreatic cancer will need to be evaluated in prospective studies.

Temporal Variability of Oral Microbiota over 10 Months and the Implications for Future Epidemiologic Studies.

Background: Few studies have prospectively evaluated the association between oral microbiota and health outcomes. Precise estimates of the intrasubject microbial metric stability will allow better study planning. Therefore, we conducted a study to evaluate the temporal variability of oral microbiota.Methods: Forty individuals provided six oral samples using the OMNIgene ORAL kit and Scope mouthwash oral rinses approximately every two months over 10 months. DNA was extracted using the QIAsymphony and the V4 region of the 16S rRNA gene was amplified and sequenced using the MiSeq. To estimate temporal variation, we calculated intraclass correlation coefficients (ICCs) for a variety of metrics and examined stability after clustering samples into distinct community types using Dirichlet multinomial models (DMMs).Results: The ICCs for the alpha diversity measures were high, including for number of observed bacterial species [0.74; 95% confidence interval (CI): 0.65-0.82 and 0.79; 95% CI: 0.75-0.94] from OMNIgene ORAL and Scope mouthwash, respectively. The ICCs for the relative abundance of the top four phyla and beta diversity matrices were lower. Three clusters provided the best model fit for the DMM from the OMNIgene ORAL samples, and the probability of remaining in a specific cluster was high (59.5%-80.7%).Conclusions: The oral microbiota appears to be stable over time for multiple metrics, but some measures, particularly relative abundance, were less stable.Impact: We used this information to calculate stability-adjusted power calculations that will inform future field study protocols and experimental analytic designs. Cancer Epidemiol Biomarkers Prev; 27(5); 594-600. ©2018 AACR.

Whole exome sequencing reveals a C-terminal germline variant in CEBPA-associated acute myeloid leukemia: 45-year follow up of a large family.

Familial acute myeloid leukemia is rare and linked to germline mutations in RUNX1, GATA2 or CCAAT/enhancer binding protein-α (CEBPA). We re-evaluated a large family with acute myeloid leukemia originally seen at NIH in 1969. We used whole exome sequencing to study this family, and conducted in silico bioinformatics analysis, protein structural modeling and laboratory experiments to assess the impact of the identified CEBPA Q311P mutation. Unlike most previously identified germline mutations in CEBPA, which were N-terminal frameshift mutations, we identified a novel Q311P variant that was located in the C-terminal bZip domain of C/EBPα. Protein structural modeling suggested that the Q311P mutation alters the ability of the CEBPA dimer to bind DNA. Electrophoretic mobility shift assays showed that the Q311P mu-tant had attenuated binding to DNA, as predicted by the protein modeling. Consistent with these findings, we found that the Q311P mutation has reduced transactivation, consistent with a loss-of-function mutation. From 45 years of follow up, we observed incomplete penetrance (46%) of CEBPA Q311P. This study of a large multi-generational pedigree reveals that a germline mutation in the C-terminal bZip domain can alter the ability of C/EBP-α to bind DNA and reduces transactivation, leading to acute myeloid leukemia.

Hoyeraal-Hreidarsson syndrome caused by a germline mutation in the TEL patch of the telomere protein TPP1.

Germline mutations in telomere biology genes cause dyskeratosis congenita (DC), an inherited bone marrow failure and cancer predisposition syndrome. DC is a clinically heterogeneous disorder diagnosed by the triad of dysplastic nails, abnormal skin pigmentation, and oral leukoplakia; Hoyeraal-Hreidarsson syndrome (HH), a clinically severe variant of DC, also includes cerebellar hypoplasia, immunodeficiency, and intrauterine growth retardation. Approximately 70% of DC cases are associated with a germline mutation in one of nine genes, the products of which are all involved in telomere biology. Using exome sequencing, we identified mutations in Adrenocortical Dysplasia Homolog (ACD) (encoding TPP1), a component of the telomeric shelterin complex, in one family affected by HH. The proband inherited a deletion from his father and a missense mutation from his mother, resulting in extremely short telomeres and a severe clinical phenotype. Characterization of the mutations revealed that the single-amino-acid deletion affecting the TEL patch surface of the TPP1 protein significantly compromises both telomerase recruitment and processivity, while the missense mutation in the TIN2-binding region of TPP1 is not as clearly deleterious to TPP1 function. Our results emphasize the critical roles of the TEL patch in proper stem cell function and demonstrate that TPP1 is the second shelterin component (in addition to TIN2) to be implicated in DC.