Citations

The development and performance of two mass spectrometry (MS) workflows for the intraoperative diagnosis of isocitrate dehydrogenase (IDH) mutations in glioma is implemented by independent teams at Mayo Clinic, Jacksonville, and Huashan Hospital, Shanghai. The infiltrative nature of gliomas makes rapid diagnosis necessary to guide the extent of surgical resection of central nervous system (CNS) tumors. The combination of tissue biopsy and MS analysis used here satisfies this requirement. The key feature of both described methods is the use of tandem MS to measure the oncometabolite 2-hydroxyglutarate (2HG) relative to endogenous glutamate (Glu) to characterize the presence of mutant tumor. The experiments i) provide IDH mutation status for individual patients and ii) demonstrate a strong correlation of 2HG signals with tumor infiltration. The measured ratio of 2HG to Glu correlates with IDH-mutant (IDH-mut) glioma ( P < 0.0001) in the tumor core data of both teams. Despite using different ionization methods and different mass spectrometers, comparable performance in determining IDH mutations from core tumor biopsies was achieved with sensitivities, specificities, and accuracies all at 100%. None of the 31 patients at Mayo Clinic or the 74 patients at Huashan Hospital were misclassified when analyzing tumor core biopsies. Robustness of the methodology was evaluated by postoperative re-examination of samples. Both teams noted the presence of high concentrations of 2HG at surgical margins, supporting future use of intraoperative MS to monitor for clean surgical margins. The power of MS diagnostics is shown in resolving contradictory clinical features, e.g., in distinguishing gliosis from IDH-mut glioma.

Abstract Extrachromosomal DNA (ecDNA) drives the evolution of cancer cells by conferring survival advantages over normal cells. However, the functional importance of ecDNA replication and maintenance is yet to be evaluated, and the molecular components that contributed to these processes remain largely unknown. Here, using CRISPR-C (circularization of genes and chromosomes by CRISPR) technology and synthetic circular DNA mimicking ecDNA identified in tumors, we generated ecDNA-positive (ecDNA+) cell models. Using these models, we identified the proteins involved in ecDNA replication and maintenance and found DNA damage response (DDR)-related pathways were significantly enriched. We further demonstrated that ecDNA replication led to increased DDR and genome instability and DDR was also essential for ecDNA maintenance, consistent with observations that ecDNA+ tumors exhibited higher DDR scores than ecDNA-negative (ecDNA–) tumors in patients. ATM, MDC1, and CHK2, but not ATR, were activated in ecDNA+ cells compared to ecDNA– cells. Moreover, ecDNA+ cells were more sensitive to ATM and CHK2 inhibition. We further found that type II topoisomerase TOP2B and DNA ligase III (LIG3) were critical regulators for ecDNA replication and maintenance and related DDR. In summary, our findings of the reciprocal interactions between ecDNA and DDR provide new insights into detection and therapy for ecDNA+ tumors.

Abstract Objective Transformer-based language models are prevailing in the clinical domain due to their excellent performance on clinical NLP tasks. The generalizability of those models is usually ignored during the model development process. This study evaluated the generalizability of CancerBERT, a Transformer-based clinical NLP model, along with classic machine learning models, i.e., conditional random field (CRF), bi-directional long short-term memory CRF (BiLSTM-CRF), across different clinical institutes through a breast cancer phenotype extraction task. Materials and methods Two clinical corpora of breast cancer patients were collected from the electronic health records from the University of Minnesota (UMN) and Mayo Clinic (MC), and annotated following the same guideline. We developed three types of NLP models (i.e., CRF, BiLSTM-CRF and CancerBERT) to extract cancer phenotypes from clinical texts. We evaluated the generalizability of models on different test sets with different learning strategies (model transfer vs locally trained). The entity coverage score was assessed with their association with the model performances. Results We manually annotated 200 and 161 clinical documents at UMN and MC. The corpora of the two institutes were found to have higher similarity between the target entities than the overall corpora. The CancerBERT models obtained the best performances among the independent test sets from two clinical institutes and the permutation test set. The CancerBERT model developed in one institute and further fine-tuned in another institute achieved reasonable performance compared to the model developed on local data (micro-F1: 0.925 vs 0.932). Conclusions The results indicate the CancerBERT model has superior learning ability and generalizability among the three types of clinical NLP models for our named entity recognition task. It has the advantage to recognize complex entities, e.g., entities with different labels.

PURPOSEThe advancement of natural language processing (NLP) has promoted the use of detailed textual data in electronic health records (EHRs) to support cancer research and to facilitate patient care. In this review, we aim to assess EHR for cancer research and patient care by using the Minimal Common Oncology Data Elements (mCODE), which is a community-driven effort to define a minimal set of data elements for cancer research and practice. Specifically, we aim to assess the alignment of NLP-extracted data elements with mCODE and review existing NLP methodologies for extracting said data elements.METHODSPublished literature studies were searched to retrieve cancer-related NLP articles that were written in English and published between January 2010 and September 2020 from main literature databases. After the retrieval, articles with EHRs as the data source were manually identified. A charting form was developed for relevant study analysis and used to categorize data including four main topics: metadata, EHR data and targeted cancer types, NLP methodology, and oncology data elements and standards.RESULTSA total of 123 publications were selected finally and included in our analysis. We found that cancer research and patient care require some data elements beyond mCODE as expected. Transparency and reproductivity are not sufficient in NLP methods, and inconsistency in NLP evaluation exists.CONCLUSIONWe conducted a comprehensive review of cancer NLP for research and patient care using EHRs data. Issues and barriers for wide adoption of cancer NLP were identified and discussed.

Abstract Objective Accurate extraction of breast cancer patients’ phenotypes is important for clinical decision support and clinical research. This study developed and evaluated cancer domain pretrained CancerBERT models for extracting breast cancer phenotypes from clinical texts. We also investigated the effect of customized cancer-related vocabulary on the performance of CancerBERT models. Materials and Methods A cancer-related corpus of breast cancer patients was extracted from the electronic health records of a local hospital. We annotated named entities in 200 pathology reports and 50 clinical notes for 8 cancer phenotypes for fine-tuning and evaluation. We kept pretraining the BlueBERT model on the cancer corpus with expanded vocabularies (using both term frequency-based and manually reviewed methods) to obtain CancerBERT models. The CancerBERT models were evaluated and compared with other baseline models on the cancer phenotype extraction task. Results All CancerBERT models outperformed all other models on the cancer phenotyping NER task. Both CancerBERT models with customized vocabularies outperformed the CancerBERT with the original BERT vocabulary. The CancerBERT model with manually reviewed customized vocabulary achieved the best performance with macro F1 scores equal to 0.876 (95% CI, 0.873–0.879) and 0.904 (95% CI, 0.902–0.906) for exact match and lenient match, respectively. Conclusions The CancerBERT models were developed to extract the cancer phenotypes in clinical notes and pathology reports. The results validated that using customized vocabulary may further improve the performances of domain specific BERT models in clinical NLP tasks. The CancerBERT models developed in the study would further help clinical decision support.

Background: Among patients (pts) with coronary artery disease (CAD) who are candidates for coronary artery bypass graft surgery (CABG), a reduced ejection fraction (EF) identifies a high-risk group. The profile of pre-operative, intra-operative and post-operative factors contributing to 30 day (d) mortality in pts with low EF undergoing CABG has not been well characterized. Methods: The STICH trial enrolled 2136 pts with CAD amenable to CABG with an LVEF < 35% who were treated with medical therapy (MED) alone, MED with CABG, or with MED plus CABG and surgical ventricular resconstruction (SVR). Multivariable logistic regression models utilizing extensive baseline, intra-operative and post-operative care variables were developed to determine factors associated with mortality at 30d in a landmark analysis among pts surviving the first 36 hours (h). Results: Among those enrolled, 1460 pts received CABG; 673 (46.1%) on pump, 160 (11.0%) off pump, 132 (9.0%) with a mitral valve procedure (MVR), 393 (26.9%) with SVR, and 102 (7.0%) received CABG with SVR and MVR. Mortality in the operating room (OR) was 0.3 %, within the first 36h was 1.0%, and at 30d was 5.1%. The rate of in-hospital acute myocardial infarction was 0.8%, stroke 1.6%, mediastinitis 1.7%, worsening renal insufficiency 8.4%, and return to OR was 6.9%. Inotropes were used in 43.6% and intra-aortic balloon pump in 18.6%. At 36h, 148 pts (10.2%) remained intubated. Among pts who survived to 36h, a robust multivariable model (Table, c-statistic 0.89) was developed to predict 30d mortality. A prolonged intubation time (>36h) was the strongest marker of 30d mortality (odds ratio 7.4, 95% CI 3.8, 14.5, p<0.0001). Conclusions: Among pts with ischemic heart failure, CABG can be performed with a low rate of complications and intra-operative mortality. Prolonged intubation time (>36h) after CABG is an integrated clincal marker that may serve to identify those pts with the least likelihood of survival to 30d.