A study to assess the efficacy of enasidenib and risk-adapted addition of azacitidine in newly diagnosed IDH2-mutant AML.

ABSTRACT: Enasidenib (ENA) is an inhibitor of isocitrate dehydrogenase 2 (IDH2) approved for the treatment of patients with IDH2-mutant relapsed/refractory acute myeloid leukemia (AML). In this phase 2/1b Beat AML substudy, we applied a risk-adapted approach to assess the efficacy of ENA monotherapy for patients aged ≥60 years with newly diagnosed IDH2-mutant AML in whom genomic profiling demonstrated that mutant IDH2 was in the dominant leukemic clone. Patients for whom ENA monotherapy did not induce a complete remission (CR) or CR with incomplete blood count recovery (CRi) enrolled in a phase 1b cohort with the addition of azacitidine. The phase 2 portion assessing the overall response to ENA alone demonstrated efficacy, with a composite complete response (cCR) rate (CR/CRi) of 46% in 60 evaluable patients. Seventeen patients subsequently transitioned to phase 1b combination therapy, with a cCR rate of 41% and 1 dose-limiting toxicity. Correlative studies highlight mechanisms of clonal elimination with differentiation therapy as well as therapeutic resistance. This study demonstrates both efficacy of ENA monotherapy in the upfront setting and feasibility and applicability of a risk-adapted approach to the upfront treatment of IDH2-mutant AML. This trial is registered at www.clinicaltrials.gov as #NCT03013998.

Application of a Domain-specific BERT for Detection of Speech Recognition Errors in Radiology Reports.

Purpose: To develop radiology domain-specific bidirectional encoder representations from transformers (BERT) models that can identify speech recognition (SR) errors and suggest corrections in radiology reports. Materials and Methods: A pretrained BERT model, Clinical BioBERT, was further pretrained on a corpus of 114 008 radiology reports between April 2016 and August 2019 that were retrospectively collected from two hospitals. Next, the model was fine-tuned on a training dataset of generated insertion, deletion, and substitution errors, creating Radiology BERT. This model was retrospectively evaluated on an independent dataset of radiology reports with generated errors (n = 18 885) and on unaltered report sentences (n = 2000) and prospectively evaluated on true clinical SR errors (n = 92). Correction Radiology BERT was separately trained to suggest corrections for detected deletion and substitution errors. Area under the receiver operating characteristic curve (AUC) and bootstrapped 95% CIs were calculated for each evaluation dataset. Results: Radiology-specific BERT had AUC values of >.99 (95% CI: >0.99, >0.99), 0.94 (95% CI: 0.93, 0.94), 0.98 (95% CI: 0.98, 0.98), and 0.97 (95% CI: 0.97, 0.97) for detecting insertion, deletion, substitution, and all errors, respectively, on the independently generated test set. Testing on unaltered report impressions revealed a sensitivity of 82% (28 of 34; 95% CI: 70%, 93%) and specificity of 88% (1521 of 1728; 95% CI: 87%, 90%). Testing on prospective SR errors showed an accuracy of 75% (69 of 92; 95% CI: 65%, 83%). Finally, the correct word was the top suggestion for 45.6% (475 of 1041; 95% CI: 42.5%, 49.3%) of errors. Conclusion: Radiology-specific BERT models fine-tuned on generated errors were able to identify SR errors in radiology reports and suggest corrections.Keywords: Computer Applications, Technology Assessment Supplemental material is available for this article. © RSNA, 2022See also the commentary by Abajian and Cheung in this issue.

Clinical language search algorithm from free-text: facilitating appropriate imaging.

BACKGROUND: The comprehensiveness and maintenance of the American College of Radiology (ACR) Appropriateness Criteria (AC) makes it a unique resource for evidence-based clinical imaging decision support, but it is underutilized by clinicians. To facilitate the use of imaging recommendations, we develop a natural language processing (NLP) search algorithm that automatically matches clinical indications that physicians write into imaging orders to appropriate AC imaging recommendations. METHODS: We apply a hybrid model of semantic similarity from a sent2vec model trained on 223 million scientific sentences, combined with term frequency inverse document frequency features. AC documents are ranked based on their embeddings' cosine distance to query. For model testing, we compiled a dataset of simulated simple and complex indications for each AC document (n = 410) and another with clinical indications from randomly sampled radiology reports (n = 100). We compare our algorithm to a custom google search engine. RESULTS: On the simulated indications, our algorithm ranked ground truth documents as top 3 for 98% of simple queries and 85% of complex queries. Similarly, on the randomly sampled radiology report dataset, the algorithm ranked 86% of indications with a single match as top 3. Vague and distracting phrases present in the free-text indications were main sources of errors. Our algorithm provides more relevant results than a custom Google search engine, especially for complex queries. CONCLUSIONS: We have developed and evaluated an NLP algorithm that matches clinical indications to appropriate AC guidelines. This approach can be integrated into imaging ordering systems for automated access to guidelines.

Development and web deployment of an automated neuroradiology MRI protocoling tool with natural language processing.

BACKGROUND: A systematic approach to MRI protocol assignment is essential for the efficient delivery of safe patient care. Advances in natural language processing (NLP) allow for the development of accurate automated protocol assignment. We aim to develop, evaluate, and deploy an NLP model that automates protocol assignment, given the clinician indication text. METHODS: We collected 7139 spine MRI protocols (routine or contrast) and 990 head MRI protocols (routine brain, contrast brain, or other) from a single institution. Protocols were split into training (n = 4997 for spine MRI; n = 839 for head MRI), validation (n = 1071 for spine MRI, fivefold cross-validation used for head MRI), and test (n = 1071 for spine MRI; n = 151 for head MRI) sets. fastText and XGBoost were used to develop 2 NLP models to classify spine and head MRI protocols, respectively. A Flask-based web app was developed to be deployed via Heroku. RESULTS: The spine MRI model had an accuracy of 83.38% and a receiver operator characteristic area under the curve (ROC-AUC) of 0.8873. The head MRI model had an accuracy of 85.43% with a routine brain protocol ROC-AUC of 0.9463 and contrast brain protocol ROC-AUC of 0.9284. Cancer, infectious, and inflammatory related keywords were associated with contrast administration. Structural anatomic abnormalities and stroke/altered mental status were indicative of routine spine and brain MRI, respectively. Error analysis revealed increasing the sample size may improve performance for head MRI protocols. A web version of the model is provided for demonstration and deployment. CONCLUSION: We developed and web-deployed two NLP models that accurately predict spine and head MRI protocol assignment, which could improve radiology workflow efficiency.

Preclinical evaluation of the Hsp90 inhibitor SNX-5422 in ibrutinib resistant CLL.

B-cell receptor (BCR) antagonists such as the BTK inhibitor ibrutinib have proven to effectively target chronic lymphocytic leukemia (CLL) tumor cells, leading to impressive response rates in these patients. However patients do still relapse on ibrutinib, and the progressive disease is often quite aggressive requiring immediate treatment. Several strategies are being pursued to treat patients who relapse on ibrutinib therapy. As the most common form of relapse is the development of a mutant form of BTK which limits ibrutinib binding, agents which lead to degradation of the BTK protein are a promising strategy. Our study explores the efficacy of the Hsp90 inhibitor, SNX-5422, in CLL. The SNX Hsp90 inhibitor was effective in primary CLL cells, as well as B-cell lines expressing either BTK wild type or C481 mutant BTK, which has been identified as the primary resistance mechanism to ibrutinib in CLL patients. Furthermore the combination of SNX-5422 and ibrutinib provided a remarkable in vivo survival benefit in the Eµ-TCL1 mouse model of CLL compared to the vehicle or single agent groups (51 day median survival in the vehicle and ibrutinib groups versus 100 day median survival in the combination). We report here preclinical data suggesting that the Hsp90 inhibitor SNX-5422, which has been pursued in clinical trials in both solid tumor and hematological malignancies, is a potential therapy for ibrutinib resistant CLL.

Domain specific word embeddings for natural language processing in radiology.

BACKGROUND: There has been increasing interest in machine learning based natural language processing (NLP) methods in radiology; however, models have often used word embeddings trained on general web corpora due to lack of a radiology-specific corpus. PURPOSE: We examined the potential of Radiopaedia to serve as a general radiology corpus to produce radiology specific word embeddings that could be used to enhance performance on a NLP task on radiological text. MATERIALS AND METHODS: Embeddings of dimension 50, 100, 200, and 300 were trained on articles collected from Radiopaedia using a GloVe algorithm and evaluated on analogy completion. A shallow neural network using input from either our trained embeddings or pre-trained Wikipedia 2014 + Gigaword 5 (WG) embeddings was used to label the Radiopaedia articles. Labeling performance was evaluated based on exact match accuracy and Hamming loss. The McNemar's test with continuity and the Benjamini-Hochberg correction and a 5×2 cross validation paired two-tailed t-test were used to assess statistical significance. RESULTS: For accuracy in the analogy task, 50-dimensional (50-D) Radiopaedia embeddings outperformed WG embeddings on tumor origin analogies (p < 0.05) and organ adjectives (p < 0.01) whereas WG embeddings tended to outperform on inflammation location and bone vs. muscle analogies (p < 0.01). The two embeddings had comparable performance on other subcategories. In the labeling task, the Radiopaedia-based model outperformed the WG based model at 50, 100, 200, and 300-D for exact match accuracy (p < 0.001, p < 0.001, p < 0.01, and p < 0.05, respectively) and Hamming loss (p < 0.001, p < 0.001, p < 0.01, and p < 0.05, respectively). CONCLUSION: We have developed a set of word embeddings from Radiopaedia and shown that they can preserve relevant medical semantics and augment performance on a radiology NLP task. Our results suggest that the cultivation of a radiology-specific corpus can benefit radiology NLP models in the future.

Targeting the C481S Ibrutinib-Resistance Mutation in Bruton's Tyrosine Kinase Using PROTAC-Mediated Degradation.

Inhibition of Bruton's tyrosine kinase (BTK) with the irreversible inhibitor ibrutinib has emerged as a transformative treatment option for patients with chronic lymphocytic leukemia (CLL) and other B-cell malignancies, yet >80% of CLL patients develop resistance due to a cysteine to serine mutation at the site covalently bound by ibrutinib (C481S). Currently, an effective treatment option for C481S patients exhibiting relapse to ibrutinib does not exist, and these patients have poor outcomes. To address this, we have developed a PROteolysis TArgeting Chimera (PROTAC) that induces degradation of both wild-type and C481S mutant BTK. We selected a lead PROTAC, MT-802, from several candidates on the basis of its potency to induce BTK knockdown. MT-802 recruits BTK to the cereblon E3 ubiquitin ligase complex to trigger BTK ubiquitination and degradation via the proteasome. MT-802 binds fewer off-target kinases than ibrutinib does and retains an equivalent potency (>99% degradation at nanomolar concentrations) against wild-type and C481S BTK. In cells isolated from CLL patients with the C481S mutation, MT-802 is able to reduce the pool of active, phosphorylated BTK whereas ibrutinib cannot. Collectively, these data provide a basis for further preclinical study of BTK PROTACs as a novel strategy for treatment of C481S mutant CLL.

The long noncoding RNA, treRNA, decreases DNA damage and is associated with poor response to chemotherapy in chronic lymphocytic leukemia.

The study of long noncoding RNAs (lncRNAs) is an emerging area of cancer research, in part due to their ability to serve as disease biomarkers. However, few studies have investigated lncRNAs in chronic lymphocytic leukemia (CLL). We have identified one particular lncRNA, treRNA, which is overexpressed in CLL B-cells. We measured transcript expression in 144 CLL patient samples and separated samples into high or low expression of treRNA relative to the overall median. We found that high expression of treRNA is significantly associated with shorter time to treatment. High treRNA also correlates with poor prognostic indicators such as unmutated IGHV and high ZAP70 protein expression. We validated these initial findings in samples collected in a clinical trial comparing the nucleoside analog fludarabine alone or in combination with the alkylating agent cyclophosphamide in untreated CLL samples collected prior to starting therapy (E2997). High expression of treRNA was independently prognostic for shorter progression free survival in patients receiving fludarabine plus cyclophosphamide. Given these results, in order to study the role of treRNA in DNA damage response we generated a model cell line system where treRNA was over-expressed in the human B-CLL cell line OSU-CLL. Relative to the vector control line, there was less cell death in OSU-CLL over-expressing treRNA after exposure to fludarabine and mafosfamide, due in part to a reduction in DNA damage. Therefore, we suggest that treRNA is a novel biomarker in CLL associated with aggressive disease and poor response to chemotherapy through enhanced protection against cytotoxic mediated DNA damage.

Hsp90 inhibition increases SOCS3 transcript and regulates migration and cell death in chronic lymphocytic leukemia.

Epigenetic or transcriptional silencing of important tumor suppressors has been described to contribute to cell survival and tumorigenesis in chronic lymphocytic leukemia (CLL). Using gene expression microarray analysis, we found that thousands of genes are repressed more than 2-fold in CLL compared to normal B cells; however therapeutic approaches to reverse this have been limited in CLL. Following treatment with the Hsp90 inhibitor 17-DMAG, a significant number of these repressed genes were significantly re-expressed. One of the genes significantly repressed in CLL and up-regulated by 17-DMAG was suppressor of cytokine signaling 3, (SOCS3). SOCS3 has been shown to be silenced in solid tumors as well as myeloid leukemia; however little is known about the regulation in CLL. We found that 17-DMAG induces expression of SOCS3 by via the activation of p38 signaling, and subsequently inhibits AKT and STAT3 phosphorylation resulting in downstream effects on cell migration and survival. We therefore suggest that SOCS3 is an important signaling protein in CLL, and Hsp90 inhibitors represent a novel approach to target transcriptional repression in B cell lymphoproliferative disorders which exhibit a substantial degree of gene repression.

Characterization of a new chronic lymphocytic leukemia cell line for mechanistic in vitro and in vivo studies relevant to disease.

Studies of chronic lymphocytic leukemia (CLL) have yielded substantial progress, however a lack of immortalized cell lines representative of the primary disease has hampered a full understanding of disease pathogenesis and development of new treatments. Here we describe a novel CLL cell line (OSU-CLL) generated by EBV transformation, which displays a similar cytogenetic and immunophenotype observed in the patient's CLL (CD5 positive with trisomy 12 and 19). A companion cell line was also generated from the same patient (OSU-NB). This cell line lacked typical CLL characteristics, and is likely derived from the patient's normal B cells. In vitro migration assays demonstrated that OSU-CLL exhibits migratory properties similar to primary CLL cells whereas OSU-NB has significantly reduced ability to migrate spontaneously or towards chemokine. Microarray analysis demonstrated distinct gene expression patterns in the two cell lines, including genes on chromosomes 12 and 19, which is consistent with the cytogenetic profile in this cell line. Finally, OSU-CLL was readily transplantable into NOG mice, producing uniform engraftment by three weeks with leukemic cells detectable in the peripheral blood spleen and bone marrow. These studies describe a new CLL cell line that extends currently available models to study gene function in this disease.

The proteasome inhibitor carfilzomib functions independently of p53 to induce cytotoxicity and an atypical NF-κB response in chronic lymphocytic leukemia cells.

PURPOSE: The proteasome consists of chymotrypsin-like (CT-L), trypsin-like, and caspase-like subunits that cleave substrates preferentially by amino acid sequence. Proteasomes mediate degradation of regulatory proteins of the p53, Bcl-2, and nuclear factor-κB (NF-κB) families that are aberrantly active in chronic lymphocytic leukemia (CLL). CLL remains an incurable disease, and new treatments are especially needed in the relapsed/refractory setting. We therefore investigated the effects of the proteasome inhibitor carfilzomib (CFZ) in CLL cells. EXPERIMENTAL DESIGN: Tumor cells from CLL patients were assayed in vitro using immunoblotting, real-time polymerase chain reaction, and electrophoretic mobility shift assays. In addition, a p53 dominant-negative construct was generated in a human B-cell line. RESULTS: Unlike bortezomib, CFZ potently induces apoptosis in CLL patient cells in the presence of human serum. CLL cells have significantly lower basal CT-L activity compared to normal B and T cells, although activity is inhibited similarly in T cells versus CLL. Co-culture of CLL cells on stroma protected from CFZ-mediated cytotoxicity; however, PI3K inhibition significantly diminished this stromal protection. CFZ-mediated cytotoxicity in leukemic B cells is caspase-dependent and occurs irrespective of p53 status. In CLL cells, CFZ promotes atypical activation of NF-κB evidenced by loss of cytoplasmic IκBα, phosphorylation of IκBα, and increased p50/p65 DNA binding, without subsequent increases in canonical NF-κB target gene transcription. CONCLUSIONS: Together, these data provide new mechanistic insights into the activity of CFZ in CLL and support phase I investigation of CFZ in this disease.