Artificial intelligence-aided clinical annotation of a large multi-cancer genomic dataset

To accelerate cancer research that correlates biomarkers with clinical endpoints, methods are needed to ascertain outcomes from electronic health records at scale. Here, we train deep natural language processing (NLP) models to extract outcomes for participants with any of 7 solid tumors in a precision oncology study. Outcomes are extracted from 305,151 imaging reports for 13,130 patients and 233,517 oncologist notes for 13,511 patients, including patients with 6 additional cancer types. NLP models recapitulate outcome annotation from these documents, including the presence of cancer, progression/worsening, response/improvement, and metastases, with excellent discrimination (AUROC > 0.90). Models generalize to cancers excluded from training and yield outcomes correlated with survival. Among patients receiving checkpoint inhibitors, we confirm that high tumor mutation burden is associated with superior progression-free survival ascertained using NLP. Here, we show that deep NLP can accelerate annotation of molecular cancer datasets with clinically meaningful endpoints to facilitate discovery.

Functional antibody and T cell immunity following SARS-CoV-2 infection, including by variants of concern, in patients with cancer: the CAPTURE study

Patients with cancer have higher COVID-19 morbidity and mortality. Here we present the prospective CAPTURE study, integrating longitudinal immune profiling with clinical annotation. Of 357 patients with cancer, 118 were SARS-CoV-2 positive, 94 were symptomatic and 2 died of COVID-19. In this cohort, 83% patients had S1-reactive antibodies and 82% had neutralizing antibodies against wild type SARS-CoV-2, whereas neutralizing antibody titers against the Alpha, Beta and Delta variants were substantially reduced. S1-reactive antibody levels decreased in 13% of patients, whereas neutralizing antibody titers remained stable for up to 329 days. Patients also had detectable SARS-CoV-2-specific T cells and CD4+ responses correlating with S1-reactive antibody levels, although patients with hematological malignancies had impaired immune responses that were disease and treatment specific, but presented compensatory cellular responses, further supported by clinical recovery in all but one patient. Overall, these findings advance the understanding of the nature and duration of the immune response to SARS-CoV-2 in patients with cancer.

Functional antibody and T-cell immunity following SARS-CoV-2 infection, including by variants of concern, in patients with cancer: the CAPTURE study

Patients with cancer have higher COVID-19 morbidity and mortality. Here we present the prospective CAPTURE study (NCT03226886) integrating longitudinal immune profiling with clinical annotation. Of 357 patients with cancer, 118 were SARS-CoV-2-positive, 94 were symptomatic and 2 patients died of COVID-19. In this cohort, 83% patients had S1-reactive antibodies, 82% had neutralizing antibodies against WT, whereas neutralizing antibody titers (NAbT) against the Alpha, Beta, and Delta variants were substantially reduced. Whereas S1-reactive antibody levels decreased in 13% of patients, NAbT remained stable up to 329 days. Patients also had detectable SARS-CoV-2-specific T cells and CD4+ responses correlating with S1-reactive antibody levels, although patients with hematological malignancies had impaired immune responses that were disease and treatment-specific, but presented compensatory cellular responses, further supported by clinical. Overall, these findings advance the understanding of the nature and duration of immune response to SARS-CoV-2 in patients with cancer.

Genomic Classification and Clinical Outcome in Rhabdomyosarcoma: A Report From an International Consortium

PURPOSE Rhabdomyosarcoma is the most common soft tissue sarcoma of childhood. Despite aggressive therapy, the 5-year survival rate for patients with metastatic or recurrent disease remains poor, and beyond PAX-FOXO1 fusion status, no genomic markers are available for risk stratification. We present an international consortium study designed to determine the incidence of driver mutations and their association with clinical outcome. PATIENTS AND METHODS Tumor samples collected from patients enrolled on Children's Oncology Group trials (1998-2017) and UK patients enrolled on malignant mesenchymal tumor and RMS2005 (1995-2016) trials were subjected to custom-capture sequencing. Mutations, indels, gene deletions, and amplifications were identified, and survival analysis was performed. RESULTS DNA from 641 patients was suitable for analyses. A median of one mutation was found per tumor. In FOXO1 fusion-negative cases, mutation of any RAS pathway member was found in > 50% of cases, and 21% had no putative driver mutation identified. BCOR (15%), NF1 (15%), and TP53 (13%) mutations were found at a higher incidence than previously reported and TP53 mutations were associated with worse outcomes in both fusion-negative and FOXO1 fusion-positive cases. Interestingly, mutations in RAS isoforms predominated in infants < 1 year (64% of cases). Mutation of MYOD1 was associated with histologic patterns beyond those previously described, older age, head and neck primary site, and a dismal survival. Finally, we provide a searchable companion database ( ClinOmics ), containing all genomic variants, and clinical annotation including survival data. CONCLUSION This is the largest genomic characterization of clinically annotated rhabdomyosarcoma tumors to date and provides prognostic genetic features that refine risk stratification and will be incorporated into prospective trials.

MutViz 2.0: visual analysis of somatic mutations and the impact of mutational signatures on selected genomic regions

Patterns of somatic single nucleotide variants observed in human cancers vary widely between different tumor types. They depend not only on the activity of diverse mutational processes, such as exposure to ultraviolet light and the deamination of methylated cytosines, but largely also on the sequence content of different genomic regions on which these processes act. With MutViz (http://gmql.eu/mutviz/), we have presented a user-friendly web tool for the identification of mutation enrichments that offers preloaded mutations from public datasets for a variety of cancer types, well organized within an effective database architecture. Somatic mutation patterns can be visually and statistically analyzed within arbitrary sets of small, user-provided genomic regions, such as promoters or collections of transcription factor binding sites. Here, we present MutViz 2.0, a largely extended and consolidated version of the tool: we took into account the immediate (trinucleotide) sequence context of mutations, improved the representation of clinical annotation of tumor samples and devised a method for signature refitting on limited genomic regions to infer the contribution of individual mutational processes to the mutation patterns observed in these regions. We described both the features of MutViz 2.0, concentrating on the novelties, and the substantial re-engineering of the cloud-based architecture.

Moving with the Times: The Health Science Alliance (HSA) Biobank, Pathway to Sustainability

Human biobanks are recognised as vital components of translational research infrastructure. With the growth in personalised and precision medicine, and the associated expansion of biomarkers and novel therapeutics under development, it is critical that researchers can access a strong collection of patient biospecimens, annotated with clinical data. Biobanks globally are undertaking transformation of their operating models in response to changing research needs; transition from a ‘classic’ model representing a largely retrospective collection of pre-defined specimens to a more targeted, prospective collection model, although there remains a research need for both models to co-exist. Here we introduce the Health Science Alliance (HSA) Biobank, established in 2012 as a classic biobank, now transitioning to a hybrid operational model. Some of the past and current challenges encountered are discussed including clinical annotation, specimen utilisation and biobank sustainability, along with the measures the HSA Biobank is taking to address these challenges. We describe new directions being explored, going beyond traditional specimen collection into areas involving bioimages, microbiota and live cell culture. The HSA Biobank is working in collaboration with clinicians, pathologists and researchers, piloting a sustainable, robust platform with the potential to integrate future needs.

Adaptive immunity to SARS-CoV-2 in cancer patients: The CAPTURE study

SUMMARYThere is a pressing need to characterise the nature, extent and duration of immune response to SARS-CoV-2 in cancer patients and inform risk-reduction strategies and preserve cancer outcomes. CAPTURE is a prospective, longitudinal cohort study of cancer patients and healthcare workers (HCWs) integrating longitudinal immune profiling and clinical annotation. We evaluated 529 blood samples and 1051 oronasopharyngeal swabs from 144 cancer patients and 73 HCWs and correlated with >200 clinical variables. In patients with solid cancers and HCWs, S1-reactive and neutralising antibodies to SARS-CoV-2 were detectable five months post-infection. SARS-CoV-2-specific T-cell responses were detected, and CD4+ T-cell responses correlated with S1 antibody levels. Patients with haematological malignancies had impaired but partially compensated immune responses. Overall, cancer stage, disease status, and therapies did not correlate with immune responses. These findings have implications for understanding individual risks and potential effectiveness of SARS-CoV-2 vaccination in the cancer population.

ClassifyCNV: a tool for clinical annotation of copy-number variants

Copy-number variants (CNVs) are an important part of human genetic variation. They can be benign or can play a role in human disease by creating dosage imbalances and disrupting genes and regulatory elements. Accurate identification and clinical annotation of CNVs is essential, however, manual evaluation of individual CNVs by clinicians is challenging on a large scale. Here, we present ClassifyCNV, an easy-to-use tool that implements the 2019 ACMG classification guidelines to assess CNV pathogenicity. ClassifyCNV uses genomic coordinates and CNV type as input and reports a clinical classification for each variant, a classification score breakdown, and a list of genes of potential importance for variant interpretation. We validate ClassifyCNV’s performance using a set of known clinical CNVs and a set of manually evaluated variants. ClassifyCNV matches the pathogenicity category for 81% of manually evaluated variants with the significance of the remaining pathogenic and benign variants automatically determined as uncertain, requiring a further evaluation by a clinician. ClassifyCNV facilitates the implementation of the latest ACMG guidelines in high-throughput CNV analysis, is suitable for integration into NGS analysis pipelines, and can decrease time to diagnosis. The tool is available at https://github.com/Genotek/ClassifyCNV.