Genomic Classification Research Articles

Genome-wide rare variant score associates with morphological subtypes of autism spectrum disorder

Defining different genetic subtypes of autism spectrum disorder (ASD) can enable the prediction of developmental outcomes. Based on minor physical and major congenital anomalies, we categorize 325 Canadian children with ASD into dysmorphic and nondysmorphic subgroups. We develop a method for calculating a patient-level, genome-wide rare variant score (GRVS) from whole-genome sequencing (WGS) data. GRVS is a sum of the number of variants in morphology-associated coding and non-coding regions, weighted by their effect sizes. Probands with dysmorphic ASD have a significantly higher GRVS compared to those with nondysmorphic ASD (P = 0.03). Using the polygenic transmission disequilibrium test, we observe an over-transmission of ASD-associated common variants in nondysmorphic ASD probands (P = 2.9 × 10−3). These findings replicate using WGS data from 442 ASD probands with accompanying morphology data from the Simons Simplex Collection. Our results provide support for an alternative genomic classification of ASD subgroups using morphology data, which may inform intervention protocols.

The path towards consensus genome classification of diffuse large B-cell lymphoma for use in clinical practice.

Diffuse large B-cell lymphoma (DLBCL) is a widely heterogeneous disease in presentation, treatment response and outcome that results from a broad biological heterogeneity. Various stratification approaches have been proposed over time but failed to sufficiently capture the heterogeneous biology and behavior of the disease in a clinically relevant manner. The most recent DNA-based genomic subtyping studies are a major step forward by offering a level of refinement that could serve as a basis for exploration of personalized and targeted treatment for the years to come. To enable consistent trial designs and allow meaningful comparisons between studies, harmonization of the currently available knowledge into a single genomic classification widely applicable in daily practice is pivotal. In this review, we investigate potential avenues for harmonization of the presently available genomic subtypes of DLBCL inspired by consensus molecular classifications achieved for other malignancies. Finally, suggestions for laboratory techniques and infrastructure required for successful clinical implementation are described.

Identification and in-silico characterization of splice-site variants from a large cardiogenetic national registry.

Splice-site variants in cardiac genes may predispose carriers to potentially lethal arrhythmias. To investigate, we screened 1315 probands and first-degree relatives enrolled in the Canadian Hearts in Rhythm Organization (HiRO) registry. 10% (134/1315) of patients in the HiRO registry carry variants within 10 base-pairs of the intron-exon boundary with 78% (104/134) otherwise genotype negative. These 134 probands were carriers of 57 unique variants. For each variant, American College of Medical Genetics and Genomics (ACMG) classification was revisited based on consensus between nine in silico tools. Due in part to the in silico algorithms, seven variants were reclassified from the original report, with the majority (6/7) downgraded. Our analyses predicted 53% (30/57) of variants to be likely/pathogenic. For the 57 variants, an average of 9 tools were able to score variants within splice sites, while 6.5 tools responded for variants outside these sites. With likely/pathogenic classification considered a positive outcome, the ACMG classification was used to calculate sensitivity/specificity of each tool. Among these, Combined Annotation Dependent Depletion (CADD) had good sensitivity (93%) and the highest response rate (131/134, 98%), dbscSNV was also sensitive (97%), and SpliceAI was the most specific (64%) tool. Splice variants remain an important consideration in gene elusive inherited arrhythmia syndromes. Screening for intronic variants, even when restricted to the ±10 positions as performed here may improve genetic testing yield. We compare 9 freely available in silico tools and provide recommendations regarding their predictive capabilities. Moreover, we highlight several novel cardiomyopathy-associated variants which merit further study.

Naryaviridae, Nenyaviridae, and Vilyaviridae: three new families of single-stranded DNA viruses in the phylum Cressdnaviricota.

The phylum Cressdnaviricota includes viruses with circular single-stranded DNA (ssDNA) genomes and icosahedral capsids. These viruses display global environmental distribution and infect diverse eukaryotic hosts, including animals, plants, and fungi. Here, we report on the formal creation of two new orders, Rivendellvirales and Rohanvirales, and three new families, Naryaviridae, Nenyaviridae, and Vilyaviridae, of ssDNA viruses associated with protozoan parasites belonging to the genera Entamoeba and Giardia. We describe a sequence-based taxonomic framework, which was used to classify 60 ssDNA viruses into 12 genera (with 18 species) within the family Vilyaviridae; four genera (with five species) within the family Naryaviridae; and five genera (with six species) within the family Nenyaviridae. We also highlight the challenges associated with the classification of chimeric virus genomes, such as those in the families Naryaviridae and Nenyaviridae, where the replicationinitiation and capsid protein genes have undergone several independent non-orthologous replacements. The described taxonomic changes have been ratified by the International Committee on Taxonomy of Viruses (ICTV)and expand the phylum Cressdnaviricota to eight orders and 11 families.

Exome sequencing for structurally normal fetuses-yields and ethical issues.

The yield of chromosomal microarray analysis (CMA) is well established in structurally normal fetuses (0.4-1.4%). We aimed to determine the incremental yield of exome sequencing (ES) in this population. From February 2017 to April 2022, 1,526 fetuses were subjected to ES; 482 of them were structurally normal (31.6%). Only pathogenic and likely pathogenic (P/LP) variants, per the American College of Medical Genetics and Genomics (ACMG) classification, were reported. Additionally, ACMG secondary findings relevant to childhood were reported. Four fetuses (4/482; 0.8%) had P/LP variants indicating a moderate to severe disease in ATP7B, NR2E3, SPRED1 and FGFR3, causing Wilson disease, Enhanced S-cone syndrome, Legius and Muenke syndromes, respectively. Two fetuses had secondary findings, in RET and DSP. Our data suggest that offering only CMA for structurally normal fetuses may provide false reassurance. Prenatal ES mandates restrictive analysis and careful management combined with pre and post-test genetic counseling.

Genomic Classification of HER2-Positive Patients With 80-Gene and 70-Gene Signatures Identifies Diversity in Clinical Outcomes With HER2-Targeted Neoadjuvant Therapy.

PURPOSEThe prospective Neoadjuvant Breast Registry Symphony Trial compared the 80-gene molecular subtyping signature with clinical assessment by immunohistochemistry and/or fluorescence in situ hybridization in predicting pathologic complete response (pCR) and 5-year outcomes in patients with early-stage breast cancer.METHODSStandard-of-care neoadjuvant chemotherapy combined with trastuzumab or trastuzumab plus pertuzumab was given to patients with human epidermal growth factor receptor 2 (HER2)–positive tumors (n = 295). pCR was the primary end point, with secondary end points of distant metastasis-free survival and overall survival at 5 years.RESULTSAmong clinically defined HER2-positive (cHER2) tumors, the 80-gene assay identified 29.5% (87 of 295) as Luminal-Type (cHER2/gLuminal), 14.9% (44 of 295) as Basal-Type (cHER2/gBasal), and 55.6% (164 of 295) as HER2-Type (cHER2/genomically classified as HER2 [gHER2]). Patients with cHER2/gHER2 tumors had a higher pCR rate (61.6%) compared with non-gHER2 tumors (26.7%; P < .001). Dual targeting for cHER2/gHER2 tumors yielded a higher pCR rate (75%) compared with those treated with single HER2-targeted therapy (54%; P = .006). For cHER2/gBasal tumors, the 42.9% pCR rate observed with dual targeting was not different from that with trastuzumab alone (46.4%; P = .830). Among those with cHER2/gBasal tumors, 5-year distant metastasis-free survival (68.6%; 95% CI, 49.1 to 81.9) was significantly worse than in patients with cHER2/gLuminal tumors (88.9%; 95% CI, 78.0 to 94.6) and cHER2/gHER2 tumors (87.4%; 95% CI, 80.2 to 92.2; P = .010), with similar corresponding overall survival differences.CONCLUSIONThe 80-gene assay identified meaningful genomic diversity in patients with cHER2 disease. Patients with cHER2/gHER2 tumors, who benefitted most from dual HER2-targeted therapy, accounted for approximately half of the cHER2 cohort. Genomically Luminal tumors had low pCR rates but good 5-year outcomes. cHER2/gBasal tumors derived no benefit from dual therapy and had significantly worse 5-year prognosis; these patients merit special consideration in future trials.

Comparative Proteomic Analysis of Liver Tissues and Serum in db/db Mice.

Background and Aims: Non-alcoholic fatty liver disease (NAFLD) affects one-quarter of individuals worldwide. Liver biopsy, as the current reliable method for NAFLD evaluation, causes low patient acceptance because of the nature of invasive sampling. Therefore, sensitive non-invasive serum biomarkers are urgently needed. Results: The serum gene ontology (GO) classification and Kyoto encyclopedia of genes and genomes (KEGG) analysis revealed the DEPs enriched in pathways including JAK-STAT and FoxO. GO analysis indicated that serum DEPs were mainly involved in the cellular process, metabolic process, response to stimulus, and biological regulation. Hepatic proteomic KEGG analysis revealed the DEPs were mainly enriched in the PPAR signaling pathway, retinol metabolism, glycine, serine, and threonine metabolism, fatty acid elongation, biosynthesis of unsaturated fatty acids, glutathione metabolism, and steroid hormone biosynthesis. GO analysis revealed that DEPs predominantly participated in cellular, biological regulation, multicellular organismal, localization, signaling, multi-organism, and immune system processes. Protein-protein interaction (PPI) implied diverse clusters of the DEPs. Besides, the paralleled changes of the common upregulated and downregulated DEPs existed in both the liver and serum were validated in the mRNA expression of NRP1, MUP3, SERPINA1E, ALPL, and ALDOB as observed in our proteomic screening. Methods: We conducted hepatic and serum proteomic analysis based on the leptin-receptor-deficient mouse (db/db), a well-established diabetic mouse model with overt obesity and NAFLD. The results show differentially expressed proteins (DEPs) in hepatic and serum proteomic analysis. A parallel reaction monitor (PRM) confirmed the authenticity of the selected DEPs. Conclusion: These results are supposed to offer sensitive non-invasive serum biomarkers for diabetes and NAFLD.

Assessing the clinical utility of protein structural analysis in genomic variant classification: experiences from a diagnostic laboratory

BackgroundThe widespread clinical application of genome-wide sequencing has resulted in many new diagnoses for rare genetic conditions, but testing regularly identifies variants of uncertain significance (VUS). The remarkable rise in the amount of genomic data has been paralleled by a rise in the number of protein structures that are now publicly available, which may have clinical utility for the interpretation of missense and in-frame insertions or deletions.MethodsWithin a UK National Health Service genomic medicine diagnostic laboratory, we investigated the number of VUS over a 5-year period that were evaluated using protein structural analysis and how often this analysis aided variant classification.ResultsWe found 99 novel missense and in-frame variants across 67 genes that were initially classified as VUS by our diagnostic laboratory using standard variant classification guidelines and for which further analysis of protein structure was requested. Evidence from protein structural analysis was used in the re-assessment of 64 variants, of which 47 were subsequently reclassified as pathogenic or likely pathogenic and 17 remained as VUS. We identified several case studies where protein structural analysis aided variant interpretation by predicting disease mechanisms that were consistent with the observed phenotypes, including loss-of-function through thermodynamic destabilisation or disruption of ligand binding, and gain-of-function through de-repression or escape from proteasomal degradation.ConclusionsWe have shown that using in silico protein structural analysis can aid classification of VUS and give insights into the mechanisms of pathogenicity. Based on our experience, we propose a generic evidence-based workflow for incorporating protein structural information into diagnostic practice to facilitate variant classification.

Clinical and biological markers for risk-stratification of T1 high-grade non-muscle invasive bladder cancer.

To summarize the prognostic and predictive role of current clinical and biological markers in patients with T1 high-grade (T1HG) nonmuscle invasive bladder cancer (NMIBC). Classical clinico-pathologic markers such as age, tumor size, focality, and location as well as the presence of concomitant carcinoma in situ, lymphovascular invasion, and histological variants at the time of transurethral resection (TUR) should be used in the risk-stratification of T1HG to improve patients' selection for early aggressive treatment. pathological T1 substaging has shown to predict disease progression and response to intravesical therapy, and should therefore be reported in the pathological assessment to improve clinical decision-making. Urinary inducible cytokines measured at different time points during Bacillus Calmette-Guerin therapy may be used to predict response to treatment, while urinary mRNA-based biomarkers may be of value to select patients for repeated TUR (reTUR). The advent of genomic classification in NMIBC and that of immune markers may improve current risk-stratification tools and pave the way toward personalized treatment. The role of clinico-pathologic variables in the risk-stratification of T1HG NMIBC remains unaltered, despite insufficient. Urinary biomarkers and tissue-based immune markers hold the promise to revolutionize the paradigm of risk-stratification due to their potential role in predicting response to intravesical and systemic immunotherapy. However, to date, none of the investigated biomarkers is used in clinical practice to risk-stratify T1HG patients due to the lack of external and/or prospective validations.

EDITORIAL COMMENT

The evolution of risk stratifications for prostate cancer reflects advances in anatomic (radiographic), pathologic, and molecular (genomic classifications) understandings. These parameters, combined with traditional “bedside” evaluations of clinical T stage, and PSA values, allow for better sharing of clinical decisions with patients. In this issue of the Gold Journal, investigators MUSIC investigators (Michigan Urological Surgery Improvement Collaborative) sought to improve patient selection criteria for PLND in patients with favorable and unfavorable intermediate risk (F-IR, U-IR) disease by analyzing over 6,800 men who underwent RALP and pelvic lymph node dissection (PLND).

DeepSMILE: Contrastive self-supervised pre-training benefits MSI and HRD classification directly from H&E whole-slide images in colorectal and breast cancer.

We propose a Deep learning-based weak label learning method for analyzing whole slide images (WSIs) of Hematoxylin and Eosin (H&E) stained tumor tissue not requiring pixel-level or tile-level annotations using Self-supervised pre-training and heterogeneity-aware deep Multiple Instance LEarning (DeepSMILE). We apply DeepSMILE to the task of Homologous recombination deficiency (HRD) and microsatellite instability (MSI) prediction. We utilize contrastive self-supervised learning to pre-train a feature extractor on histopathology tiles of cancer tissue. Additionally, we use variability-aware deep multiple instance learning to learn the tile feature aggregation function while modeling tumor heterogeneity. For MSI prediction in a tumor-annotated and color normalized subset of TCGA-CRC (n=360 patients), contrastive self-supervised learning improves the tile supervision baseline from 0.77 to 0.87 AUROC, on par with our proposed DeepSMILE method. On TCGA-BC (n=1041 patients) without any manual annotations, DeepSMILE improves HRD classification performance from 0.77 to 0.81 AUROC compared to tile supervision with either a self-supervised or ImageNet pre-trained feature extractor. Our proposed methods reach the baseline performance using only 40% of the labeled data on both datasets. These improvements suggest we can use standard self-supervised learning techniques combined with multiple instance learning in the histopathology domain to improve genomic label classification performance with fewer labeled data.

Diagnosis of Brain Tumor Using Light Weight Deep Learning Model with Fine-Tuning Approach

Brain cancer is a rare and deadly disease with a slim chance of survival. One of the most important tasks for neurologists and radiologists is to detect brain tumors early. Recent claims have been made that computer-aided diagnosis-based systems can diagnose brain tumors by employing magnetic resonance imaging (MRI) as a supporting technology. We propose transfer learning approaches for a deep learning model to detect malignant tumors, such as glioblastoma, using MRI scans in this study. This paper presents a deep learning-based approach for brain tumor identification and classification using the state-of-the-art object detection framework YOLO (You Only Look Once). The YOLOv5 is a novel object detection deep learning technique that requires limited computational architecture than its competing models. The study used the Brats 2021 dataset from the RSNA-MICCAI brain tumor radio genomic classification. The dataset has images annotated from RSNA-MICCAI brain tumor radio genomic competition dataset using the make sense an AI online tool for labeling dataset. The preprocessed data is then divided into testing and training for the model. The YOLOv5 model provides a precision of 88 percent. Finally, our model is tested across the whole dataset, and it is concluded that it is able to detect brain tumors successfully.

A Genetic Folding Strategy Based Support Vector Machine to Optimize Lung Cancer Classification.

Cancer is defined as an abnormal growth of human cells classified into benign and malignant. The site makes further classification of cancers of initiation and genomic underpinnings. Lung cancer displays extreme heterogeneity, making genomic classification vital for future targeted therapies. Especially considering lung cancers account for 1.76 million deaths worldwide annually. However, tumors do not always correlate to cancer as they can be benign, severely dysplastic (pre-cancerous), or malignant (cancerous). Lung cancer presents with ambiguous symptoms, thus is difficult to diagnose and is detected later compared to other cancers. Diagnosis relies heavily on radiology and invasive procedures. Different models developed employing Artificial Intelligence (AI), and Machine Learning (ML) have been used to classify various cancers. In this study, the authors propose a Genetic Folding Strategy (GFS) based model to predict lung cancer from a lung cancer dataset. We developed and implemented GF to improve Support Vector Machines (SVM) classification kernel functions and used it to classify lung cancer. We developed and implemented GF to improve SVM classification kernel functions and used it to classify lung cancer. Classification performance evaluations and comparisons between the authors' GFS model and three SVM kernels, linear, polynomial and radial basis function, were conducted thoroughly on real lung cancer datasets. While using GFS in classifying lung cancer, the authors obtained an accuracy of 96.2%. This is the highest current accuracy compared to other kernels.

P1026: COMPARATIVE GENOMIC PROFILING OF MYELOPROLIFERATIVE NEOPLASMS PRESENTING WITH AND WITHOUT SPLANCHNIC VEIN THROMBOSIS

Background: Splanchnic vein thrombosis (SVT) is the initial manifestation in 5% of myeloproliferative neoplasms (MPN). MPN patients presenting with SVT are usually younger, have less abnormal blood counts and lower JAK2V617F allele burden than other MPN patients. However, it is unknown whether they have a different genomic background. Aims: To describe the clinical and genomic profile of MPN presenting with SVT and to compare it with a matched group of MPN without SVT. Methods: Samples of 190 MPN patients were selected: 64 MPN presenting with SVT (SVT+) and 126 MPN without SVT (SVT-). SVT+ and SVT- were matched according to sex, age, MPN clinical subtype and driver mutation. Next Generation Sequencing was performed using Myeloid Solution by Sophia Genetics, which includes 30 genes recurrently mutated in MPN. Variants were selected using Sophia DDM software and were subsequently classified into 5 categories: 5 (definitely pathogenic), 4 (likely pathogenic), 3 (uncertain), 2 (likely benign), and 1 (benign). Only variants included into 4 and 5 categories were taken into account. Patients were classified according to the genomic classification proposed by Grinfeld and colleagues. Results: One hundred and three patients were women (54.2%) and 87 patients were men (45.8%), without significant differences between SVT+ and SVT- groups. Median age was 44 years (range: 15-85) and 50 years (range: 15-87) in SVT+ and SVT-, respectively (p not significant). Regarding MPN clinical subtype in SVT+ group, 39 (60.9%), 15 (23.4%) and 10 (15.6%) corresponded to Polycythemia Vera (PV), Essential Thrombocythemia (ET) and MPN unclassifiable, respectively. In the SVT- group, 75 (59.5%) and 51 (40.5%) corresponded to PV and ET, respectively. Genomic classification in the SVT+ and SVT- groups is shown in Table 1. The proportion of patients with high molecular risk (TP53 disruption/aneuploidy, chromatin/spliceosome mutation, or homozygous JAK2 mutation) was 14.1% and 26.2% for SVT+ and SVT-, respectively (p=0.06). Median number of pathogenic mutations was 1 (range: 1-4) in both SVT+ and SVT- groups. Pathogenic mutations in TET2, DNMT3A and ASXL1 were present in 10%, 3.7% and 3.7%, respectively, without significant differences between both groups. With a mean follow-up of 10.1 years for cases and 9.7 years for controls, 27 patients died (13 SVT+ and 14 SVT-). Median survival was 24 years and not reached for SVT+ and SVT-, respectively (p=0.1). On multivariate analysis, SVT+ group showed a higher risk of death (HR: 2.4, 95%CI: 1.1-5.5, p=0.03) after correction by age, sex and genomic classification. Transformation to myelofibrosis and acute leukemia was low and similar in both groups: 4.7% vs. 5.6% for myelofibrosis in SVT+ and SVT-, respectively (p not significant); and 0% vs. 2.4% for acute leukemia (p not significant). On multivariate analysis, patients with TP53 disruption/aneuploidy, chromatin/spliceosome mutation, or homozygous JAK2 showed a higher risk of disease progression to myelofibrosis or acute leukemia (HR: 14.5, 95%CI: 3.1-68.7, p=0.001) regardless of the presence of SVT at diagnosis, age and sex. Image:Summary/Conclusion: MPN presenting with SVT and MPN-matched controls showed a low molecular complexity. Genomic classification was helpful in identifying a small proportion of MPN patients presenting with SVT who are at high risk of disease progression. SVT at MPN diagnosis is associated with a higher risk of death independent of sex, age and genomic classification.

PB1847: ACUTE MYELOID LEUKEMIA – IS IT MORE BESIDES GENETIC ABNORMALITIES?

Background: Acute myeloid leukaemia (AML) still carries a poor prognosis, despite novel aproaches. Age and previous myelodysplastic syndrome (MDS) are important risk factors. Environmental factors: professional benzene exposure, residential background and exposure to chemotherapy (CT)/radiotherapy (RT) have also been implicated. Aims: To evaluate environmental factors of AML patients, its genetic characteristics and overall survival (OS), in a tertiary centre. Methods: We retrospectively revised newly diagnosed AML in our centre between 2018 - 2020. Clinical and laboratory data were obtained by clinical files review Results: We included 235 patients, 47.2% (n=111) male. Median age 69 (19-97) years. (67 [28.5%]≤60 years old. – YOUNG and 168 [71.5%] with >60 years old - OLD). The most common WHO 2016 category was AML with myelodysplasia-related changes (MRC) (37.0%; n=87), followed by AML NOS (21.3%; n=50); 15 (6.4%) patients had acute promyelocytic leukaemia with PML―RARA, 21 (8.9%) therapy-related myeloid neoplasm and 17 (7.2%) post-myeloproliferative neoplasm (post-NMP) AML. The main category within YOUNG patients was AML with mutated NPM1 (19.4%) and in the OLD group was AML-MRC (45.2%). Genomic classification (GC) (Papaemmanuil, 2016) was available for 117 cases (49.8%): AML with no detected driver mutations (n=28; 23.9%); AML with mutated chromatin, RNA-splicing genes, or both (n=26; 22.2%). Considering YOUNG vs OLD patients, the main category was AML with mutated NPM1 (25%) vs AML with no detected driver mutations (32.3%). Regarding ELN risk stratification: 20.0% were low-risk, 27.1% intermediate risk and 52.9% high-risk category. OS for total cohort was 4.2 months, with 177 events (75.3% of included patients). The YOUNG patients had a median OS of 18,7 months while the OLD had median OS of 2,6 months, with P<0,001 (HR 2,31). Ninety nine (42.1%) patients lived in industrial regions. We found no correlation between living place and dysplasia, even after adjusting for age, ELN risk or GC. We found an apparent association of WHO subtype with living place (p=0.013), as post-MPN AML was more common in non-industrial (10.3 vs 3.0%) and AML with mutated NPM1 in industrial areas (11.1 vs 7.3%), but they were lost after adjusting for age. There was no OS difference between industrial and non-industrial places even after adjusting for age. Regarding professional exposure (available for 85 patients), 7.1% were exposed to benzene, 2.4% to agricultural activities and 31.8% to other chemicals, 58.8% with no significant exposure, with no differences in WHO classification, GC nor age. Smoking appeared associated with longer OS (6.6 vs 34.9%; p=0.026; HR 0.53), but this effect is lost after adjusting for age, as smoking was more common in YOUNG (65.7 vs 25.0%). There was no correlation with AML subtype or risk. Concerning other risk factors like alcohol consumption, obesity, family history of cancer, previous solid neoplasm, previous chemotherapy or previous radiotherapy none of them had impact in AML subtype, ELN risk or OS (p=NS). Summary/Conclusion: In our cohort, the majority of patients were over 60 years old and presented with AML-MRC, as described in the literature. Concerning GC, the dominant category was AML with no detected driver mutations. Looking at environmental factors, in our cohort only residential background correlated with AML subtype, despite non-significance impact in prognosis.

An enhanced Genetic Folding algorithm for prostate and breast cancer detection.

Cancer’s genomic complexity is gradually increasing as we learn more about it. Genomic classification of various cancers is crucial in providing oncologists with vital information for targeted therapy. Thus, it becomes more pertinent to address issues of patient genomic classification. Prostate cancer is a cancer subtype that exhibits extreme heterogeneity. Prostate cancer contributes to 7.3% of new cancer cases worldwide, with a high prevalence in males. Breast cancer is the most common type of cancer in women and the second most significant cause of death from cancer in women. Breast cancer is caused by abnormal cell growth in the breast tissue, generally referred to as a tumour. Tumours are not synonymous with cancer; they can be benign (noncancerous), pre-malignant (pre-cancerous), or malignant (cancerous). Fine-needle aspiration (FNA) tests are used to biopsy the breast to diagnose breast cancer. Artificial Intelligence (AI) and machine learning (ML) models are used to diagnose with varying accuracy. In light of this, we used the Genetic Folding (GF) algorithm to predict prostate cancer status in a given dataset. An accuracy of 96% was obtained, thus being the current highest accuracy in prostate cancer diagnosis. The model was also used in breast cancer classification with a proposed pipeline that used exploratory data analysis (EDA), label encoding, feature standardization, feature decomposition, log transformation, detect and remove the outliers with Z-score, and the BAGGINGSVM approach attained a 95.96% accuracy. The accuracy of this model was then assessed using the rate of change of PSA, age, BMI, and filtration by race. We discovered that integrating the rate of change of PSA and age in our model raised the model’s area under the curve (AUC) by 6.8%, whereas BMI and race had no effect. As for breast cancer classification, no features were removed.

Genome Analysis of Goose-Origin Astroviruses Causing Fatal Gout in Shanghai, China Reveals One of Them Belonging to a Novel Type Is a Recombinant Strain.

Since 2014, a goose-origin astroviruses disease, which is characterized by urate precipitation in viscera, has rapidly spread to major commercial goose provinces leading to huge economic losses in the poultry industry of China. In March 2020, a goose farm locates in Shanghai, China, where there was no goose astroviruses (GAstVs) infection reported before, experienced an outbreak of gout disease in geese. The etiological investigation was carried out by virus metagenomics and bacterial culture and two GAstVs strains, designated as CHSH01 and CHSH02, were determined. Their complete genomes were measured to 7,154 and 7,330 nt in length, excludingthe poly(A) tail, respectively, and had different genomic features and classifications. CHSH01 shared a very low sequence identity with other strains in terms of not only the complete genome but also different ORFs. Phylogenetic analysis showed CHSH02 belonged to GAstV-2, which was the predominant species in the geese with gout in China according to the previous study. Meanwhile, CHSH01 strain displayed low identity with other AstVs, and phylogenetic and recombination analysis suggested that CHSH01 belonging to a novel type was a recombinant strain, one parent strain of which was an AstV determined from a bar-headed goose (a kind of migrant bird). Moreover, the primary epidemiological investigation showed that the two strains were prevalent in the same goose farm and co-infection occurred. These findings arise the potential cross-species transmission of CHSH01 between domestic and wild fowl.

Abstract LB113: Genomic classification to refine prognosis in clear cell renal cell carcinoma

Abstract Renal cell carcinomas (RCC) are characterized by their heterogenous clinical outcomes, and due to their indeterminate behavior and the absence of routine biomarkers, it is difficult to identify patients who are at high-risk for relapse after curative nephrectomy. To identify genomic biomarkers for clear cell RCC (ccRCC) risk-stratification we interrogated somatic mutation status of 12 RCC-relevant genes using next-generation sequencing (NGS) in tumor-normal pairs from 943 patients with matched follow up data from the Cancer Genomics of the Kidney (CAGEKID) study. We examined associations between genomically-defined patient groups, explained below, and disease-free as well as RCC-specific survival independently in two cohorts of patients (N=469 for cohort 1; 474 for cohort 2). We used the Kaplan-Meier method with log-rank tests to compare survival functions, and Cox proportional hazards models to stratify for patient stage and age to estimate association of each group with survival. RCC-specific survival was assessed with a competing-risks method to include deaths from other causes. Within these cohorts, 76.4% of patients harbored somatic mutations in VHL, the most common driver gene in ccRCC. The most commonly mutated genes within VHL-mutated tumors were PBRM1 (39.7%), SETD2 (19%), BAP1 (14.3%), and KDM5C (8.3%). Less frequently mutated genes included ATM, COL11A1, DMD, TP53, and TRRAP (~3-5%).Among VHL-driven tumors, we identified a new genomic classifier on the basis of the number of mutations in additional RCC driver genes in the panel examined. Patients were classified based on the presence of mutations only in VHL (VHL+0), those with mutations in VHL and one other driver gene (VHL+1), two other driver genes (VHL+2), and 3 or more other driver genes (VHL≥3). We observed within both cohorts that both the risk of disease recurrence as well as RCC-specific death were associated with an increased number of mutations within this classification. When stratified for patient stage and age, the hazard-ratio for 5-year disease-free survival for VHL≥3 patients was 6.69 (p=0.000212), 4.31 for VHL+2 (p=0.000862), and 2.43 for VHL+1 (p=0.035662), compared to patients with only mutations in VHL. These observations were replicated in the second patient cohort, with hazards ratios of 4.55, 2.49, and 1.40, for VHL≥3, VHL+2, and VHL+1 classified patients respectively, indicating that risk of disease recurrence increases with the number of driver mutations. Notably, tumor mutational burden (TMB) was not significantly different between the aforementioned groups, demonstrating that our classifier is independent of TMB. We created a model based on a set of 12 RCC-relevant genes, which can predict risk of relapse for the ~80% of patients with ccRCC that are VHL-driven. This classification can be defined based on a small panel of genes, making it easily applicable to the clinic, in the context of tumor or liquid biopsy analysis. Citation Format: Kate I. Glennon, Naveen S. Vasudev, Ghislaine Scelo, Michelle Wilson, Louis Letourneau, Robert Eveleigh, Nazanin Nourbehesht, Madeleine Arseneault, Antoine Paccard, Lars Egevad, Juris Viksna, Edgars Celms, Sharon M. Jackson, Behnoush Abedi-Ardekani, Anne Y. Warren, Peter J. Selby, Sebastian Trainor, Michael Kimuli, Naeem Soomro, Adebanji Adeyoju, Poulam Patel, Magdalena B. Wozniak, Ivana Holcatova, Antonin Brisuda, Vladimir Janout, Estelle Chanudet, David Zaridze, Anush Moukeria, Oxana Shangina, Lenka Foretova, Marie Navratilova, Dana Mates, Viorel Jinga, Ljiljana Bogdanovic, Bozidar Kovacevic, Anne Cambon-Thomsen, Guillaume Bourque, Alvis Brazma, Jörg Tost, Paul Brennan, Mark Lathrop, Yasser Riazalhosseini, Rosamonde E. Banks. Genomic classification to refine prognosis in clear cell renal cell carcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr LB113.

A Dual Barcoding Approach to Bacterial Strain Nomenclature: Genomic Taxonomy of Klebsiella pneumoniae Strains.

Sublineages (SLs) within microbial species can differ widely in their ecology and pathogenicity, and their precise definition is important in basic research and for industrial or public health applications. Widely accepted strategies to define SLs are currently missing, which confuses communication in population biology and epidemiological surveillance. Here, we propose a broadly applicable genomic classification and nomenclature approach for bacterial strains, using the prominent public health threat Klebsiella pneumoniae as a model. Based on a 629-gene core genome multilocus sequence typing (cgMLST) scheme, we devised a dual barcoding system that combines multilevel single linkage (MLSL) clustering and life identification numbers (LINs). Phylogenetic and clustering analyses of >7,000 genome sequences captured population structure discontinuities, which were used to guide the definition of 10 infraspecific genetic dissimilarity thresholds. The widely used 7-gene multilocus sequence typing (MLST) nomenclature was mapped onto MLSL SLs (threshold: 190 allelic mismatches) and clonal group (threshold: 43) identifiers for backwards nomenclature compatibility. The taxonomy is publicly accessible through a community-curated platform (https://bigsdb.pasteur.fr/klebsiella), which also enables external users’ genomic sequences identification. The proposed strain taxonomy combines two phylogenetically informative barcode systems that provide full stability (LIN codes) and nomenclatural continuity with previous nomenclature (MLSL). This species-specific dual barcoding strategy for the genomic taxonomy of microbial strains is broadly applicable and should contribute to unify global and cross-sector collaborative knowledge on the emergence and microevolution of bacterial pathogens.

Identifying the Best Ki-67 Cut-Off for Determining Luminal Breast Cancer Subtypes Using Immunohistochemical Analysis and PAM50 Genomic Classification

Background: A surrogate classification of breast cancer (BC) molecular subtypes based on immunohistochemistry (IHC) was established at the 13th St. Gallen International Breast Cancer Consensus (SG-BCC). The most controversial point of discussion was the difference between the luminal A and B subtypes. The Ki-67 cut-off that has been used to differentiate these BC subtypes is 14%; however, this cut-off was questioned. This study aimed to identifying the best Ki67 cut-off for determining the luminal BC by PAM50/Prosigna (NanoString Technologies, Seattle, Washington, USA). Methods: This study included females who were diagnosed with early-stage luminal BC between 2015–2020, and whose samples were subjected to genomic testing using PAM50. Results: A total of 143 samples were analysed. At the Ki-67 cut-off values of &gt;14%; a correlation of 70.6%, with a sensitivity of 79.1% and a specificity of 55.8%; and a positive predictive value of 75.8% and negative predictive value of 60.4% were observed. When the Ki-67 cut-off was increased to &gt;20%, the percentage of well-classified tumours based on IHC was 76.2%, increasing the agreement by 6.2%. The sensitivity was 93.4%, but the specificity was 46.1%. The positive predictive value was 75.2% while the negative predictive value was 80%, suggesting that IHC has a high probability of diagnosing luminal A and B. Conclusions: Increasing the Ki-67 cut-off to &gt;20% leads to a better surrogate classification based on IHC and to a higher sensitivity in classifying the luminal subtypes. The authors propose that the cut-off for Ki-67, which is an independent factor, should be globally modified to &gt;20%.