A curated census of pathogenic and likely pathogenic UTR variants and evaluation of deep learning models for variant effect prediction.

Introduction: Variants in 5' and 3' untranslated regions (UTR) contribute to rare disease. While predictive algorithms to assist in classifying pathogenicity can potentially be highly valuable, the utility of these tools is often unclear, as it depends on carefully selected training and validation conditions. To address this, we developed a high confidence set of pathogenic (P) and likely pathogenic (LP) variants and assessed deep learning (DL) models for predicting their molecular effects. Methods: 3' and 5' UTR variants documented as P or LP (P/LP) were obtained from ClinVar and refined by reviewing the annotated variant effect and reassessing evidence of pathogenicity following published guidelines. Prediction scores from sequence-based DL models were compared between three groups: P/LP variants acting though the mechanism for which the model was designed (model-matched), those operating through other mechanisms (model-mismatched), and putative benign variants. PhyloP was used to compare conservation scores between P/LP and putative benign variants. Results: 295 3' and 188 5' UTR variants were obtained from ClinVar, of which 26 3' and 68 5' UTR variants were classified as P/LP. Predictions by DL models achieved statistically significant differences when comparing modelmatched P/LP variants to both putative benign variants and modelmismatched P/LP variants, as well as when comparing all P/LP variants to putative benign variants. PhyloP conservation scores were significantly higher among P/LP compared to putative benign variants for both the 3' and 5' UTR. Discussion: In conclusion, we present a high-confidence set of P/LP 3' and 5' UTR variants spanning a range of mechanisms and supported by detailed pathogenicity and molecular mechanism evidence curation. Predictions from DL models further substantiate these classifications. These datasets will support further development and validation of DL algorithms designed to predict the functional impact of variants that may be implicated in rare disease.

The Neoantigen Landscape of the Coding and Noncoding Cancer Genome Space.

Tumor mutation burden (TMB) is a measure to predict patient responsiveness to immune checkpoint immunotherapy because with increased mutation frequency, the likelihood of a greater neoantigen burden is increased. Although neoantigen prediction tools exist, tumor neoantigen burden has not been adopted as a measure to predict immunotherapy response. With both measures, current guidelines are limited to the coding regions, but ectopic expression of sequences in the noncoding space may potentially be a source of neoantigens. A pan-cancer cohort of 574 advanced disease stage patients with whole genome and transcriptome sequencing was analyzed to report mutation burden and neoantigen counts within the coding and noncoding regions. The efficacy of tumor neoantigen burden, reported as tumor neoantigen count (TNC), including neoantigens derived from the expression of noncoding regions, compared with TMB as a predictor of response to immunotherapy for 80 patients who had received treatment, was evaluated. TMB was found to be the best predictor of response to immunotherapy, whereas expression-derived TNC from the noncoding regions did not improve prediction of response. Therefore, there is minimal benefit in extending the calculation of TNC to the noncoding space for the purposes of predicting response. However, it is likely that there is a wealth of neoantigens derived from the noncoding space that may impact patient outcomes and treatments.

Use of Treatment-Focused Tumor Sequencing to Screen for Germline Cancer Predisposition.

Next-generation sequencing assays are capable of identifying cancer patients eligible for targeted therapies and can also detect germline variants associated with increased cancer susceptibility. However, these capabilities have yet to be routinely harmonized in a single assay because of challenges with accurately identifying germline variants from tumor-only data. We have developed the Oncology and Hereditary Cancer Program targeted capture panel, which uses tumor tissue to simultaneously screen for both clinically actionable solid tumor variants and germline variants across 45 genes. Validation using 14 tumor specimens, composed of patient samples and cell lines analyzed in triplicate, demonstrated high coverage with sensitive and specific identification of single-nucleotide variants and small insertions and deletions. Average coverage across all targets remained >2000× in 198 additional patient tumor samples. Analysis of 55 formalin-fixed, paraffin-embedded tumor samples for the detection of known germline variants within a subset of cancer-predisposition genes, including one multiexon deletion, yielded a 100% detection rate, demonstrating that germline variants can be reliably detected in tumor samples using a single panel. Combining targetable somatic and actionable germline variants into a single tumor tissue assay represents a streamlined approach that can inform treatment for patients with advanced cancers as well as identify those with potential germline variants who are eligible for confirmatory testing, but would not otherwise have been identified.

The origins and genomic diversity of American Civil War Era smallpox vaccine strains.

Vaccination has transformed public health, most notably including the eradication of smallpox. Despite its profound historical importance, little is known of the origins and diversity of the viruses used in smallpox vaccination. Prior to the twentieth century, the method, source and origin of smallpox vaccinations remained unstandardised and opaque. We reconstruct and analyse viral vaccine genomes associated with smallpox vaccination from historical artefacts. Significantly, we recover viral molecules through non-destructive sampling of historical materials lacking signs of biological residues. We use the authenticated ancient genomes to reveal the evolutionary relationships of smallpox vaccination viruses within the poxviruses as a whole.

CARD 2020: antibiotic resistome surveillance with the comprehensive antibiotic resistance database.

The Comprehensive Antibiotic Resistance Database (CARD; https://card.mcmaster.ca) is a curated resource providing reference DNA and protein sequences, detection models and bioinformatics tools on the molecular basis of bacterial antimicrobial resistance (AMR). CARD focuses on providing high-quality reference data and molecular sequences within a controlled vocabulary, the Antibiotic Resistance Ontology (ARO), designed by the CARD biocuration team to integrate with software development efforts for resistome analysis and prediction, such as CARD's Resistance Gene Identifier (RGI) software. Since 2017, CARD has expanded through extensive curation of reference sequences, revision of the ontological structure, curation of over 500 new AMR detection models, development of a new classification paradigm and expansion of analytical tools. Most notably, a new Resistomes & Variants module provides analysis and statistical summary of in silico predicted resistance variants from 82 pathogens and over 100 000 genomes. By adding these resistance variants to CARD, we are able to summarize predicted resistance using the information included in CARD, identify trends in AMR mobility and determine previously undescribed and novel resistance variants. Here, we describe updates and recent expansions to CARD and its biocuration process, including new resources for community biocuration of AMR molecular reference data.