Building large-scale registries from unstructured clinical notes using a low-resource natural language processing pipeline.

Building clinical registries is an important step in clinical research and improvement of patient care quality. Natural Language Processing (NLP) methods have shown promising results in extracting valuable information from unstructured clinical notes. However, the structure and nature of clinical notes are very different from regular text that state-of-the-art NLP models are trained and tested on, and they have their own set of challenges. In this study, we propose Sentence Extractor with Keywords (SE-K), an efficient and interpretable classification approach for extracting information from clinical notes and show that it outperforms more computationally expensive methods in text classification. Following the Institutional Review Board (IRB) approval, we used SE-K and two embedding based NLP approaches (Sentence Extractor with Embeddings (SE-E) and Bidirectional Encoder Representations from Transformers (BERT)) to develop comprehensive registry of anterior cruciate ligament surgeries from 20 years of unstructured clinical data at a multi-site tertiary-care regional children's hospital. The low-resource approach (SE-K) had better performance (average AUROC of 0.94 ± 0.04) than the embedding-based approaches (SE-E: 0.93 ± 0.04 and BERT: 0.87 ± 0.09) for out of sample validation, in addition to minimum performance drop between test and out-of-sample validation. Moreover, the SE-K approach was at least six times faster (on CPU) than SE-E (on CPU) and BERT (on GPU) and provides interpretability. Our proposed approach, SE-K, can be effectively used to extract relevant variables from clinic notes to build large-scale registries, with consistently better performance compared to the more resource-intensive approaches (e.g., BERT). Such approaches can facilitate information extraction from unstructured notes for registry building, quality improvement and adverse event monitoring.

Translational and Therapeutic Evaluation of RAS-GTP Inhibition by RMC-6236 in RAS-Driven Cancers.

RAS-driven cancers comprise up to 30% of human cancers. RMC-6236 is a RAS(ON) multi-selective noncovalent inhibitor of the active, GTP-bound state of both mutant and wild-type variants of canonical RAS isoforms with broad therapeutic potential for the aforementioned unmet medical need. RMC-6236 exhibited potent anticancer activity across RAS-addicted cell lines, particularly those harboring mutations at codon 12 of KRAS. Notably, oral administration of RMC-6236 was tolerated in vivo and drove profound tumor regressions across multiple tumor types in a mouse clinical trial with KRASG12X xenograft models. Translational PK/efficacy and PK/PD modeling predicted that daily doses of 100 mg and 300 mg would achieve tumor control and objective responses, respectively, in patients with RAS-driven tumors. Consistent with this, we describe here objective responses in two patients (at 300 mg daily) with advanced KRASG12X lung and pancreatic adenocarcinoma, respectively, demonstrating the initial activity of RMC-6236 in an ongoing phase I/Ib clinical trial (NCT05379985). SIGNIFICANCE: The discovery of RMC-6236 enables the first-ever therapeutic evaluation of targeted and concurrent inhibition of canonical mutant and wild-type RAS-GTP in RAS-driven cancers. We demonstrate that broad-spectrum RAS-GTP inhibition is tolerable at exposures that induce profound tumor regressions in preclinical models of, and in patients with, such tumors.

Systematic evaluation of common natural language processing techniques to codify clinical notes.

Proper codification of medical diagnoses and procedures is essential for optimized health care management, quality improvement, research, and reimbursement tasks within large healthcare systems. Assignment of diagnostic or procedure codes is a tedious manual process, often prone to human error. Natural Language Processing (NLP) has been suggested to facilitate this manual codification process. Yet, little is known on best practices to utilize NLP for such applications. With Large Language Models (LLMs) becoming more ubiquitous in daily life, it is critical to remember, not every task requires that level of resource and effort. Here we comprehensively assessed the performance of common NLP techniques to predict current procedural terminology (CPT) from operative notes. CPT codes are commonly used to track surgical procedures and interventions and are the primary means for reimbursement. Our analysis of 100 most common musculoskeletal CPT codes suggest that traditional approaches can outperform more resource intensive approaches like BERT significantly (P-value = 4.4e-17) with average AUROC of 0.96 and accuracy of 0.97, in addition to providing interpretability which can be very helpful and even crucial in the clinical domain. We also proposed a complexity measure to quantify the complexity of a classification task and how this measure could influence the effect of dataset size on model's performance. Finally, we provide preliminary evidence that NLP can help minimize the codification error, including mislabeling due to human error.

Comprehensive evaluation of magnetic resonance imaging sequences for signal intensity based assessment of anterior cruciate ligament healing following surgical treatment.

Normalized signal intensity (SI) obtained from magnetic resonance imaging (MRI) has been used to track anterior cruciate ligament (ACL) postoperative remodeling. We aimed to assess the effect of MRI sequence (PD: proton density-weighted; T2: T2-weighted; CISS: constructive interference in steady state) on postoperative changes in healing ACLs/grafts. We hypothesized that CISS is better at detecting longitudinal SI and texture changes of the healing ACL/graft compared to the common clinical sequences (PD and T2). MR images of patients who underwent ACL surgery were evaluated and separated into groups based on surgical procedure (Bridge-Enhanced ACL Repair (BEAR; n = 50) versus ACL reconstruction (ACLR; n = 24)). CISS images showed decreasing SI across all timepoints in both the BEAR and ACLR groups (p < 0.01), PD and T2 images showed decreasing SI in the 6-to-12- and 12-to-24-month postoperative timeframes in the BEAR group (p < 0.02), and PD images additionally showed decreasing SI between 6- and 24-months postoperation in the ACLR group (p = 0.02). CISS images showed texture changes in both the BEAR and ACLR groups, showing increases in energy and decreases in entropy in the 6-to-12- and 6-to-24-month postoperative timeframes in the BEAR group (p < $\lt $ 0.04), and increases in energy, decreases in entropy, and increases in homogeneity between 6 and 24 months postoperation in the ACLR group (p < 0.04). PD images showed increases in energy and decreases in entropy between 6- and 24-months postoperation in the ACLR group (p < 0.008). Finally, CISS was estimated to require a smaller sample size than PD and T2 to detect SI differences related to postoperative remodeling.

Tumor-selective effects of active RAS inhibition in pancreatic ductal adenocarcinoma.

Broad-spectrum RAS inhibition holds the potential to benefit roughly a quarter of human cancer patients whose tumors are driven by RAS mutations. However, the impact of inhibiting RAS functions in normal tissues is not known. RMC-7977 is a highly selective inhibitor of the active (GTP-bound) forms of KRAS, HRAS, and NRAS, with affinity for both mutant and wild type (WT) variants. As >90% of human pancreatic ductal adenocarcinoma (PDAC) cases are driven by activating mutations in KRAS, we assessed the therapeutic potential of RMC-7977 in a comprehensive range of PDAC models, including human and murine cell lines, human patient-derived organoids, human PDAC explants, subcutaneous and orthotopic cell-line or patient derived xenografts, syngeneic allografts, and genetically engineered mouse models. We observed broad and pronounced anti-tumor activity across these models following direct RAS inhibition at doses and concentrations that were well-tolerated in vivo. Pharmacological analyses revealed divergent responses to RMC-7977 in tumor versus normal tissues. Treated tumors exhibited waves of apoptosis along with sustained proliferative arrest whereas normal tissues underwent only transient decreases in proliferation, with no evidence of apoptosis. Together, these data establish a strong preclinical rationale for the use of broad-spectrum RAS inhibition in the setting of PDAC.

Bi-steric mTORC1 inhibitors induce apoptotic cell death in tumor models with hyperactivated mTORC1.

The PI3K/AKT/mTOR pathway is commonly dysregulated in cancer. Rapalogs exhibit modest clinical benefit, likely owing to their lack of effects on 4EBP1. We hypothesized that bi-steric mTORC1-selective inhibitors would have greater potential for clinical benefit than rapalogs in tumors with mTORC1 dysfunction. We assessed this hypothesis in tumor models with high mTORC1 activity both in vitro and in vivo. Bi-steric inhibitors had strong growth inhibition, eliminated phosphorylated 4EBP1, and induced more apoptosis than rapamycin or MLN0128. Multiomics analysis showed extensive effects of the bi-steric inhibitors in comparison with rapamycin. De novo purine synthesis was selectively inhibited by bi-sterics through reduction in JUN and its downstream target PRPS1 and appeared to be the cause of apoptosis. Hence, bi-steric mTORC1-selective inhibitors are a therapeutic strategy to treat tumors driven by mTORC1 hyperactivation.

LigaNET: A multi-modal deep learning approach to predict the risk of subsequent anterior cruciate ligament injury after surgery.

Anterior cruciate ligament (ACL) injuries are a common cause of soft tissue injuries in young active individuals, leading to a significant risk of premature joint degeneration. Postoperative management of such injuries, in particular returning patients to athletic activities, is a challenge with immediate and long-term implications including the risk of subsequent injury. In this study, we present LigaNET, a multi-modal deep learning pipeline that predicts the risk of subsequent ACL injury following surgical treatment. Postoperative MRIs (n=1,762) obtained longitudinally between 3 to 24 months after ACL surgery from a cohort of 159 patients along with 11 non-imaging outcomes were used to train and test: 1) a 3D CNN to predict subsequent ACL injury from segmented ACLs, 2) a 3D CNN to predict injury from the whole MRI, 3) a logistic regression classifier predict injury from non-imaging data, and 4) a multi-modal pipeline by fusing the predictions of each classifier. The CNN using the segmented ACL achieved an accuracy of 77.6% and AUROC of 0.84, which was significantly better than the CNN using the whole knee MRI (accuracy: 66.6%, AUROC: 0.70; P<.001) and the non-imaging classifier (accuracy: 70.1%, AUROC: 0.75; P=.039). The fusion of all three classifiers resulted in highest classification performance (accuracy: 80.6%, AUROC: 0.89), which was significantly better than each individual classifier (P<.001). The developed multi-modal approach had similar performance in predicting the risk of subsequent ACL injury from any of the imaging sequences (P>.10). Our results demonstrate that a deep learning approach can achieve high performance in identifying patients at high risk of subsequent ACL injury after surgery and may be used in clinical decision making to improve postoperative management (e.g., safe return to sports) of ACL injured patients.

Chemical remodeling of a cellular chaperone to target the active state of mutant KRAS.

The discovery of small-molecule inhibitors requires suitable binding pockets on protein surfaces. Proteins that lack this feature are considered undruggable and require innovative strategies for therapeutic targeting. KRAS is the most frequently activated oncogene in cancer, and the active state of mutant KRAS is such a recalcitrant target. We designed a natural product-inspired small molecule that remodels the surface of cyclophilin A (CYPA) to create a neomorphic interface with high affinity and selectivity for the active state of KRASG12C (in which glycine-12 is mutated to cysteine). The resulting CYPA:drug:KRASG12C tricomplex inactivated oncogenic signaling and led to tumor regressions in multiple human cancer models. This inhibitory strategy can be used to target additional KRAS mutants and other undruggable cancer drivers. Tricomplex inhibitors that selectively target active KRASG12C or multiple RAS mutants are in clinical trials now (NCT05462717 and NCT05379985).

Disparities in cannabis use and documentation in electronic health records among children and young adults.

The legalizations of medical and recreational cannabis have generated a great deal of interest in studying the health impacts of cannabis products. Despite increases in cannabis use, its documentation during clinical visits is not yet mainstream. This lack of information hampers efforts to study cannabis's effects on health outcomes. A clear and in-depth understanding of current trends in cannabis use documentation is necessary to develop proper guidelines to screen and document cannabis use. Here we have developed and used a natural language processing pipeline to evaluate the trends and disparities in cannabis documentation. The pipeline includes a screening step to identify clinical notes with cannabis use documentation which is then fed into a BERT-based classifier to confirm positive use. This pipeline is applied to more than 23 million notes from a large cohort of 370,087 patients seen in a high-volume multi-site pediatric and young adult clinic over a period of 21 years. Our findings show a very low but growing rate of cannabis use documentation (<2%) in electronic health records with significant demographic and socioeconomic disparities in both documentation and positive use, which requires further attention.

A bi-steric mTORC1-selective inhibitor overcomes drug resistance in breast cancer.

Activation of the PI3K-mTOR pathway is central to breast cancer pathogenesis including resistance to many targeted therapies. The mTOR kinase forms two distinct complexes, mTORC1 and mTORC2, and understanding which is required for the survival of malignant cells has been limited by tools to selectively and completely impair either subcomplex. To address this, we used RMC-6272, a bi-steric molecule with a rapamycin-like moiety linked to an mTOR active-site inhibitor that displays >25-fold selectivity for mTORC1 over mTORC2 substrates. Complete suppression of mTORC1 by RMC-6272 causes apoptosis in ER+/HER2- breast cancer cell lines, particularly in those that harbor mutations in PIK3CA or PTEN, due to inhibition of the rapamycin resistant, mTORC1 substrate 4EBP1 and reduction of the pro-survival protein MCL1. RMC-6272 reduced translation of ribosomal mRNAs, MYC target genes, and components of the CDK4/6 pathway, suggesting enhanced impairment of oncogenic pathways compared to the partial mTORC1 inhibitor everolimus. RMC-6272 maintained efficacy in hormone therapy-resistant acquired cell lines and patient-derived xenografts (PDX), showed increased efficacy in CDK4/6 inhibitor treated acquired resistant cell lines versus their parental counterparts, and was efficacious in a PDX from a patient experiencing resistance to CDK4/6 inhibition. Bi-steric mTORC1-selective inhibition may be effective in overcoming multiple forms of therapy-resistance in ER+ breast cancers.

Discovery of RMC-5552, a Selective Bi-Steric Inhibitor of mTORC1, for the Treatment of mTORC1-Activated Tumors.

Hyperactivation of mTOR kinase by mutations in the PI3K/mTOR pathway or by crosstalk with other mutant cancer drivers, such as RAS, is a feature of many tumors. Multiple allosteric inhibitors of mTORC1 and orthosteric dual inhibitors of mTORC1 and mTORC2 have been developed as anticancer drugs, but their clinical utility has been limited. To address these limitations, we have developed a novel class of "bi-steric inhibitors" that interact with both the orthosteric and the allosteric binding sites in order to deepen the inhibition of mTORC1 while also preserving selectivity for mTORC1 over mTORC2. In this report, we describe the discovery and preclinical profile of the development candidate RMC-5552 and the in vivo preclinical tool compound RMC-6272. We also present evidence that selective inhibition of mTORC1 in combination with covalent inhibition of KRASG12C shows increased antitumor activity in a preclinical model of KRASG12C mutant NSCLC that exhibits resistance to KRASG12C inhibitor monotherapy.

Accurate Estimation of Total Intracranial Volume in MRI using a Multi-tasked Image-to-Image Translation Network.

Total intracranial volume (TIV) is the volume enclosed inside the cranium, inclusive of the meninges and the brain. TIV is extensively used to correct variations in inter-subject head size for the evaluation of neurodegenerative diseases. In this work, we present an automatic method to generate a TIV mask from MR images while synthesizing a CT image to be used in subsequent analysis. In addition, we propose an alternative way to obtain ground truth TIV masks using a semi-manual approach, which results in significant time savings. We train a conditional generative adversarial network (cGAN) using 2D MR slices to realize our tasks. The quantitative evaluation showed that the model was able to synthesize CT and generate TIV masks that closely approximate the reference images. This study also provides a comparison of the described method against skull stripping tools that output a mask enclosing the cranial volume, using MRI scan. In particular, highlighting the deficiencies in using such tools to approximate the volume using MRI scan.

Efficacy of a Novel Bi-Steric mTORC1 Inhibitor in Models of B-Cell Acute Lymphoblastic Leukemia.

The mechanistic target of rapamycin (mTOR) is a kinase whose activity is elevated in hematological malignancies. mTOR-complex-1 (mTORC1) phosphorylates numerous substrates to promote cell proliferation and survival. Eukaryotic initiation factor 4E (eIF4E)-binding proteins (4E-BPs) are mTORC1 substrates with an integral role in oncogenic protein translation. Current pharmacological approaches to inhibit mTORC1 activity and 4E-BP phosphorylation have drawbacks. Recently we described a series of bi-steric compounds that are potent and selective inhibitors of mTORC1, inhibiting 4E-BP phosphorylation at lower concentrations than mTOR kinase inhibitors (TOR-KIs). Here we report the activity of the mTORC1-selective bi-steric inhibitor, RMC-4627, in BCR-ABL-driven models of B-cell acute lymphoblastic leukemia (B-ALL). RMC-4627 exhibited potent and selective inhibition of 4E-BP1 phosphorylation in B-ALL cell lines without inhibiting mTOR-complex-2 (mTORC2) activity. RMC-4627 suppressed cell cycle progression, reduced survival, and enhanced dasatinib cytotoxicity. Compared to a TOR-KI compound, RMC-4627 was more potent, and its effects on cell viability were sustained after washout in vitro. Notably, a once-weekly, well tolerated dose reduced leukemic burden in a B-ALL xenograft model and enhanced the activity of dasatinib. These preclinical studies suggest that intermittent dosing of a bi-steric mTORC1-selective inhibitor has therapeutic potential as a component of leukemia regimens, and further study is warranted.

Selective inhibitors of mTORC1 activate 4EBP1 and suppress tumor growth.

The clinical benefits of pan-mTOR active-site inhibitors are limited by toxicity and relief of feedback inhibition of receptor expression. To address these limitations, we designed a series of compounds that selectively inhibit mTORC1 and not mTORC2. These 'bi-steric inhibitors' comprise a rapamycin-like core moiety covalently linked to an mTOR active-site inhibitor. Structural modification of these components modulated their affinities for their binding sites on mTOR and the selectivity of the bi-steric compound. mTORC1-selective compounds potently inhibited 4EBP1 phosphorylation and caused regressions of breast cancer xenografts. Inhibition of 4EBP1 phosphorylation was sufficient to block cancer cell growth and was necessary for maximal antitumor activity. At mTORC1-selective doses, these compounds do not alter glucose tolerance, nor do they relieve AKT-dependent feedback inhibition of HER3. Thus, in preclinical models, selective inhibitors of mTORC1 potently inhibit tumor growth while causing less toxicity and receptor reactivation as compared to pan-mTOR inhibitors.

Barriers to point-of-care ultrasound utilization during cardiac arrest in the emergency department: a regional survey of emergency physicians.

INTRODUCTION: Though point-of-care ultrasound (POCUS) is recognized as a useful diagnostic and prognostic intervention during cardiac arrest (CA), critics advise caution. The purpose of this survey study was to determine the barriers to POCUS during CA in the Emergency Department (ED). METHODS: Two survey instruments were distributed to emergency medicine (EM) attending and resident physicians at three academic centers in the South Florida. The surveys assessed demographics, experience, proficiency, attitudes and barriers. Descriptive and inferential statistics along with Item Response Theory Logistic Model and the Friedman Test with Wilcoxon Signed Rank tests were used to profile responses and rank barriers. RESULTS: 206 EM physicians were invited to participate in the survey, and 187 (91%) responded. 59% of attending physicians and 47% of resident physicians reported that POCUS is performed in all their cases of CA. 5% of attending physicians and 0% of resident physicians reported never performing POCUS during CA. The top-ranked departmental barrier for attending physicians was "No structured curriculum to educate physicians on POCUS." The top-ranked personal barriers were "I do not feel comfortable with my POCUS skills" and "I do not have sufficient time to dedicate to learning POCUS." The top-ranked barriers for resident physicians were "Time to retrieve and operate the machine" and "Chaotic milieu." CONCLUSIONS: While our study demonstrates that most attending and resident physicians utilize POCUS in CA, barriers to high-quality implementation exist. Top attending physician barriers relate to POCUS education, while the top resident physician barriers relate to logistics and the machines. Interventions to overcome these barriers might lead to optimization of POCUS performance during CA in the ED.

Shp1 Loss Enhances Macrophage Effector Function and Promotes Anti-Tumor Immunity.

Shp1, encoded by the gene Ptpn6, is a protein tyrosine phosphatase that transduces inhibitory signals downstream of immunoreceptors in many immune cell types. Blocking Shp1 activity represents an exciting potential immunotherapeutic strategy for the treatment of cancer, as Shp1 inhibition would be predicted to unleash both innate and adaptive immunity against tumor cells. Antibodies blocking the interaction between CD47 on tumor cells and SIRPα on macrophages enhance macrophage phagocytosis, show efficacy in preclinical tumor models, and are being evaluated in the clinic. Here we found that Shp1 bound to phosphorylated peptide sequences derived from SIRPα and transduced the anti-phagocytic signal, as Shp1 loss in mouse bone marrow-derived macrophages increased phagocytosis of tumor cells in vitro. We also generated a novel mouse model to evaluate the impact of global, inducible Ptpn6 deletion on anti-tumor immunity. We found that inducible Shp1 loss drove an inflammatory disease in mice that was phenotypically similar to that seen when Ptpn6 is knocked out from birth. This indicates that acute perturbation of Shp1 in vivo could drive hyperactivation of immune cells, which could be therapeutically beneficial, though at the risk of potential toxicity. In this model, we found that Shp1 loss led to robust anti-tumor immunity against two immune-rich syngeneic tumor models that are moderately inflamed though not responsive to checkpoint inhibitors, MC38 and E0771. Shp1 loss did not promote anti-tumor activity in the non-inflamed B16F10 model. The observed activity in MC38 and E0771 tumors was likely due to effects of both innate and adaptive immune cells. Following Shp1 deletion, we observed increases in intratumoral myeloid cells in both models, which was more striking in E0771 tumors. E0771 tumors also contained an increased ratio of effector to regulatory T cells following Shp1 loss. This was not observed for MC38 tumors, though we did find increased levels of IFNγ, a cytokine produced by effector T cells, in these tumors. Overall, our preclinical data suggested that targeting Shp1 may be an attractive therapeutic strategy for boosting the immune response to cancer via a mechanism involving both innate and adaptive leukocytes.

Allosteric Inhibition of SHP2 Stimulates Antitumor Immunity by Transforming the Immunosuppressive Environment.

The protein tyrosine phosphatase SHP2 binds to phosphorylated signaling motifs on regulatory immunoreceptors including PD-1, but its functional role in tumor immunity is unclear. Using preclinical models, we show that RMC-4550, an allosteric inhibitor of SHP2, induces antitumor immunity, with effects equivalent to or greater than those resulting from checkpoint blockade. In the tumor microenvironment, inhibition of SHP2 modulated T-cell infiltrates similar to checkpoint blockade. In addition, RMC-4550 drove direct, selective depletion of protumorigenic M2 macrophages via attenuation of CSF1 receptor signaling and increased M1 macrophages via a mechanism independent of CD8+ T cells or IFNγ. These dramatic shifts in polarized macrophage populations in favor of antitumor immunity were not seen with checkpoint blockade. Consistent with a pleiotropic mechanism of action, RMC-4550 in combination with either checkpoint or CSF1R blockade caused additive antitumor activity with complete tumor regressions in some mice; tumors intrinsically sensitive to SHP2 inhibition or checkpoint blockade were particularly susceptible. Our preclinical findings demonstrate that SHP2 thus plays a multifaceted role in inducing immune suppression in the tumor microenvironment, through both targeted inhibition of RAS pathway-dependent tumor growth and liberation of antitumor immune responses. Furthermore, these data suggest that inhibition of SHP2 is a promising investigational therapeutic approach. SIGNIFICANCE: Inhibition of SHP2 causes direct and selective depletion of protumorigenic M2 macrophages and promotes antitumor immunity, highlighting an investigational therapeutic approach for some RAS pathway-driven cancers.

Design and Discovery of N-(3-(2-(2-Hydroxyethoxy)-6-morpholinopyridin-4-yl)-4-methylphenyl)-2-(trifluoromethyl)isonicotinamide, a Selective, Efficacious, and Well-Tolerated RAF Inhibitor Targeting RAS Mutant Cancers: The Path to the Clinic.

Direct pharmacological inhibition of RAS has remained elusive, and efforts to target CRAF have been challenging due to the complex nature of RAF signaling, downstream of activated RAS, and the poor overall kinase selectivity of putative RAF inhibitors. Herein, we describe 15 (LXH254, Aversa, R.; et al. Int. Patent WO2014151616A1, 2014), a selective B/C RAF inhibitor, which was developed by focusing on drug-like properties and selectivity. Our previous tool compound, 3 (RAF709; Nishiguchi, G. A.; et al. J. Med. Chem. 2017, 60, 4969), was potent, selective, efficacious, and well tolerated in preclinical models, but the high human intrinsic clearance precluded further development and prompted further investigation of close analogues. A structure-based approach led to a pyridine series with an alcohol side chain that could interact with the DFG loop and significantly improved cell potency. Further mitigation of human intrinsic clearance and time-dependent inhibition led to the discovery of 15. Due to its excellent properties, it was progressed through toxicology studies and is being tested in phase 1 clinical trials.