Internal Cohort Research Articles

KMT2C mutation as a predictor of immunotherapeutic efficacy in colorectal cancer

Immunotherapy had shown good antitumor activity in a variety of solid tumors, but low benefit in CRC, so there was an urgent need to explore new biomarkers. We evaluated the role of KMT2C using publicly available data from the Cancer Genome Atlas (TCGA) and Memorial Sloan Kettering Cancer Center (MSKCC). In addition, further analysis was performed in an internal cohort. Moreover, the mutant profiles of KMT2C was analyzed in a large CRC cohort. The relationship between clinical pathologic features and KMT2C were analyzed with using the two-sided chi-squared test or the Fisher exact test. Clinicopathologic characteristics associated with overall survival using Cox regression and the Kaplan–Meier method. We found that KMT2C-mutated CRC patients in the immunotherapy cohort had significantly improved OS compared with KMT2C WT patients (P = 0.013). However, this phenomenon did not exist in non-immunotherapy cohort. Our cohort validated the value of KMT2C mutations in predicting better clinical outcomes, including ORR (P < 0.0001) and OS (P = 0.010). Meanwhile, KMT2C mutation was associated with higher tumor mutation burden, MSI score, higher levels of immune-associated T cells, neutrophil, and M1-type macrophages. Our study suggested that KMT2C mutation might be a potential positive predictor for CRC immunotherapy.

Characterization of polyamine metabolism predicts prognosis, immune profile, and therapeutic efficacy in lung adenocarcinoma patients.

Polyamine modification patterns in lung adenocarcinoma (LUAD) and their impact on prognosis, immune infiltration, and anti-tumor efficacy have not been systematically explored. Patients from The Cancer Genome Atlas (TCGA) were classified into subtypes according to polyamine metabolism-related genes using the consensus clustering method, and the survival outcomes and immune profile were compared. Meanwhile, the geneCluster was constructed according to the differentially expressed genes (DEGs) of the subtypes. Subsequently, the polyamine metabolism-related score (PMRS) system was established using the least absolute shrinkage and selection operator (LASSO) multivariate regression analysis in the TCGA training cohort (n = 245), which can be applied to characterize the prognosis. To verify the predictive performance of the PMRS, the internal cohort (n = 245) and the external cohort (n = 244) were recruited. The relationship between the PMRS and immune infiltration and antitumor responses was investigated. Two distinct patterns (C1 and C2) were identified, in which the C1 subtype presented an adverse prognosis, high CD8+ T cell infiltration, tumor mutational burden (TMB), immune checkpoint, and low tumor immune dysfunction and exclusion (TIDE). Furthermore, two geneClusters were established, and similar findings were observed. The PMRS, including three genes (SMS, SMOX, and PSMC6), was then constructed to characterize the polyamine metabolic patterns, and the patients were divided into high- and low-PMRS groups. As confirmed by the validation cohort, the high-PMRS group possessed a poor prognosis. Moreover, external samples and immunohistochemistry confirmed that the three genes were highly expressed in tumor samples. Finally, immunotherapy and chemotherapy may be beneficial to the high-PMRS group based on the immunotherapy cohorts and low half-maximal inhibitory concentration (IC50) values. We identified distinct polyamine modification patterns and established a PMRS to provide new insights into the mechanism of polyamine action and improve the current anti-tumor strategy of LUAD.

Abstract LB386: Weakly-supervised prediction of tumor infiltrating lymphocytes per high power field from colorectal cancer histopathology slides using regression transformers

Abstract Tumor infiltrating lymphocytes (TILs)-based biomarkers have emerged as a robust method for predicting treatment efficacy and outcome in colorectal cancer (CRC). However, the quantification of TILs remains a strenuous task for pathologists and is susceptible to inter-pathologist variations. Deep learning (DL) can predict cancer biomarkers directly from routine hematoxylin and eosin (H&amp;E) pathology slides, enabling the automated and consistent quantification of TILs for clinical decision-making. Our study focuses on the prediction of TILs per high-power field (HPF) on two large cohorts of CRC patients. We developed a weakly-supervised, transformer-based deep regression model to predict TILs per HPF from routine H&amp;E-stained histology slides. We trained the model using 5-fold cross-validation on a large cohort of CRC patients (n=1,738), and validated its performance on an external cohort of CRC patients (n=223). The model was evaluated using the Pearson’s correlation coefficient r. Moreover, pathologist-determined TILs per HPF was categorized into low (&amp;lt;2) and high (≥2), enabling evaluation via the area under the receiver operating characteristic curve (AUROC). Our model achieved a significant correlation coefficient of 0.60 (p&amp;lt;0.0001) for the predicted TILs per HPF in the holdout test sets of the internal cohort, and a significant correlation coefficient of 0.57 (p&amp;lt;0.0001) on the external validation cohort. Using the thresholds for low and high TILs per HPF, we found that the model accurately predicted the presence of high TILs per HPF in both the internal and external test cohorts, yielding AUROCs of 0.76 and 0.81, respectively. These findings underscore the efficacy of our deep learning model in predicting TILs per HPF from routine pathology slides. This approach holds promise for cost-effective, efficient and uniform quantification of TILs per HPF in CRC, potentially across various cancer types. Citation Format: Omar S. El Nahhas, Joseph D. Bonner, Joel K. Greenson, Daniel B. Schmolze, Lawrence Shaktah, Jonathan Salazar, Lorena Reynaga, Sidney Lindsey, Jenny Lu, Victor Moreno, Stephanie L. Schmit, Ya-Yu Tsai, Stanley R. Hamilton, Gad Rennert, Jakob N. Kather, Stephen B. Gruber. Weakly-supervised prediction of tumor infiltrating lymphocytes per high power field from colorectal cancer histopathology slides using regression transformers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB386.

Rapid Endoscopic Diagnosis of Benign Ulcerative Colorectal Diseases With an Artificial Intelligence Contextual Framework

Benign ulcerative colorectal diseases (UCDs) such as ulcerative colitis, Crohn's disease, ischemic colitis, and intestinal tuberculosis share similar phenotypes with different etiologies and treatment strategies. To accurately diagnose closely related diseases like UCDs, we hypothesize that contextual learning is critical in enhancing the ability of the artificial intelligence models to differentiate the subtle differences in lesions amidst the vastly divergent spatial contexts. White-light colonoscopy datasets of patients with confirmed UCDs and healthy controls were retrospectively collected. We developed a Multiclass Contextual Classification (MCC) model that can differentiate among the mentioned UCDs and healthy controls by incorporating the tissue object contexts surrounding the individual lesion region in a scene and spatial information from other endoscopic frames (video-level) into a unified framework. Internal and external datasets were used to validate the model's performance. Training datasets included 762 patients, and the internal and external testing cohorts included 257 patients and 293 patients, respectively. Our MCC model provided a rapid reference diagnosis on internal test sets with a high averaged area under the receiver operating characteristic curve (image-level: 0.950 and video-level: 0.973) and balanced accuracy (image-level: 76.1% and video-level: 80.8%), which was superior to junior endoscopists (accuracy: 71.8%, P < .0001) and similar to experts (accuracy: 79.7%, P= .732). The MCC model achieved an area under the receiver operating characteristic curve of 0.988 and balanced accuracy of 85.8% using external testing datasets. These results enable this model to fit in the routine endoscopic workflow, and the contextual framework to be adopted for diagnosing other closely related diseases.

Comparison of time to treatment initiation of specialty medications between an integrated health system specialty pharmacy and external specialty pharmacies.

Specialty medications are commonly dispensed through specialty pharmacies equipped to meet unique monitoring and dispensing requirements. Integrated health system specialty pharmacies (HSSPs) coordinate with health system providers to deliver specialty medications to patients and ameliorate barriers to care. However, payors may restrict specialty medication fills to specialty pharmacies external to the health system, potentially leading to delayed treatment. To compare time to treatment initiation among patients whose specialty medications were transferred to external pharmacies and patients whose medications were filled at an internal HSSP. This was a retrospective, propensity-matched cohort study examining time to treatment initiation in patients with a specialty medication referral to the University of Kentucky HealthCare Specialty Pharmacy between July 1, 2021, and July 1, 2022. Patients were classified into cohorts by receipt of dispensing services from the internal HSSP or an external specialty pharmacy. Data collected via chart review included insurance type, reason for prescription transfer, dates of service (including prescription order, transfer, and receipt of medication), and whether a prior authorization or clinical intervention was performed. Subgroup analyses were performed for patients requiring a prior authorization or clinical intervention. The Wilcoxon signed-rank test was used to assess for statistically significant differences in time to treatment initiation between cohorts. A total of 560 patients with external transfers were identified for inclusion into the study, and after exclusion criteria were applied, 408 external transfer patients were propensity matched 1:1 to 408 patients with internal fills (total n = 816). Time to treatment initiation was significantly longer in the external transfer cohort as compared with the internal fill cohort, (18 days vs 12 days; P < 0.0001). The internal fill cohort had a greater mean days from provider order to the medication being ready to fill compared with the external transfer cohort (10 days vs 6 days; P < 0.0001). The internal fill cohort had fewer mean days from the medication being ready to fill to patient receipt of the medication as compared with the external transfer cohort (2 days vs 12 days; P < 0.0001). Similar findings were observed in the subgroup analyses. Average time to treatment initiation was 6 days shorter for patients whose specialty medications were filled at this HSSP compared with externally transferred patients. Delays in therapy can cause a negative impact on patient care and disease state management, with increased concern for specialty populations. The results of this study highlight the need for continued discussion about policies that limit patient choice to in-network pharmacies.

A machine learning radiomics model based on bpMRI to predict bone metastasis in newly diagnosed prostate cancer patients.

ObjectivesTo develop and evaluate a machine learning radiomics model based on bpMRI to predict bone metastasis (BM) status in newly diagnosed prostate cancer (PCa) patients. MethodsWe retrospectively analyzed biparametric magnetic resonance imaging MRI (bpMRI) scans of PCa patients from multiple centers between January 2016 and October 2021. 348 PCa patients were recruited from two institutions for this study. The first institution contributed 284 patients, stratified and randomly divided into training and internal validation cohorts at a 7:3 ratio. The remaining 64 patients were sourced from the second institution and comprised the external validation cohort. Radiomics features were extracted from axial T2-weighted imaging (T2WI) and diffusion-weighted imaging (DWI) tumor regions. We developed the radiomics prediction model for BM in the training cohort and validated it in the internal and external validation cohorts. As a benchmark, we trained the logistic regression model with lasso feature reduction (LFR-LRM) in the training cohort and further compared it with Naive Bayes, eXtreme Gradient Boosting (XGboost), Random Forest (RF), GBDT, SVM, Adaboost, and KNN algorithms and validated in both the internal and external cohorts. The performance of several predictive models was assessed by receiver operating characteristic (ROC). ResultsThe LFR-LRM model achieved an area under the receiver operating characteristic curve (AUC) of 0.89 (95% CI: 0.822–0.974) and an accuracy of 0.828 (95% CI: 0.713–0.911). The AUC and accuracy in external validation were 0.866 (95% CI: 0.784–0.948) and 0.769 (95% CI: 0.648–0.864), respectively. The RF and XGBoost models outperformed the LFR-LRM, with AUCs of 0.907 (95% CI: 0.863–0.949) and 0.928 (95% CI: 0.882–0.974) and accuracies of 0.831 (95% CI: 0.727–0.907) and 0.884 (95% CI: 0.792–0.946). External validation for these models yielded AUCs and accuracies of 0.911 (95% CI: 0.861–0.966), 0.921 (95% CI: 0.889–0.953), and 0.846 (95% CI: 0.735–0.923) and 0.876 (95% CI: 0.771–0.945), respectively. ConclusionsThe XGboost machine learning model is more accurate than LFR-LRM for predicting BM in patients with newly confirmed PCa.

Development and validation of machine learning models to predict postoperative infarction in moyamoya disease.

Cerebral infarction is a common complication in patients undergoing revascularization surgery for moyamoya disease (MMD). Although previous statistical evaluations have identified several risk factors for postoperative brain ischemia, the ability to predict its occurrence based on these limited predictors remains inadequately explored. This study aimed to assess the feasibility of machine learning algorithms for predicting cerebral infarction after revascularization surgery in patients with MMD. This retrospective study was conducted across two centers and harnessed data from 512 patients with MMD who had undergone revascularization surgery. The patient cohort was partitioned into internal and external datasets. Using perioperative clinical data from the internal cohort, three distinct machine learning algorithms-namely the support vector machine, random forest, and light gradient-boosting machine models-were trained and cross-validated to predict the occurrence of postoperative cerebral infarction. Predictive performance validity was subsequently assessed using an external dataset. Shapley additive explanations (SHAP) analysis was conducted to augment the prediction model's transparency and to quantify the impact of each input variable on shaping both the aggregate and individual patient predictions. In the cohort of 512 patients, 33 (6.4%) experienced postrevascularization cerebral infarction. The cross-validation outcomes revealed that, among the three models, the support vector machine model achieved the largest area under the receiver operating characteristic curve (ROC-AUC) at mean ± SD 0.785 ± 0.052. Notably, during external validation, the light gradient-boosting machine model exhibited the highest accuracy at 0.903 and the largest ROC-AUC at 0.710. The top-performing prediction model utilized five input variables: postoperative serum gamma-glutamyl transpeptidase value, positive posterior cerebral artery (PCA) involvement on preoperative MRA, infarction as the rationale for surgery, presence of an infarction scar on preoperative MRI, and preoperative modified Rankin Scale score. Furthermore, the SHAP analysis identified presence of PCA involvement, infarction as the rationale for surgery, and presence of an infarction scar on preoperative MRI as positive influences on postoperative cerebral infarction. This study indicates the usefulness of employing machine learning techniques with routine perioperative data to predict the occurrence of cerebral infarction after revascularization procedures in patients with MMD.

Diet and well-being of 1st year Medical Interns during hospital placement

Hospital placement is essential training for medical interns, involving shift work and high-pressure environments. This can increase physiological and psychological stress, which may be mediated by metabolites of microbial digestion(1). Nutrients of interest include those accessible to microbial digestion and associated with altered signalling within the microbiota-gut-brain axis (MGBA)(1). Fibre is fermented by gut microbes to produce short-chain fatty acids(2) and is associated with improved psychological outcomes(3). Tryptophan, a precursor to gut-derived serotonin(2), has been negatively associated with anxiety(4). Processed foods contain food additives, excess sugars, and saturated fats that may disrupt gut homeostasis(1) and impact psychological well-being(4). Lastly, total energy intake may determine the level of substrate available for microbial fermentation(2). Therefore, this research explores how microbiota-accessible food components interact with physical and psychological well-being in a cohort of medical interns undertaking their first-year of hospital placement. Participants were healthy medical interns, during first-year hospital placement (n = 21) from the Hunter New England Local Health District, NSW, Australia. Participants completed diet and wellbeing surveys at baseline and every 2 months over a 10-month period. 24-hour diet diaries were self-recorded from participants using a mobile application (Easy Diet Diary) and analysed using AusNut and the NOVA classification system of ultra-processed foods (ULP). Wellbeing surveys include depression, anxiety, stress scale (DASS), and PROMIS survey for mental (M), physical (P), and sleep well-being. Current data represents an ‘in-progress’ of the longitudinal data collection. This study utilised Spearman correlation and Tukey’s post hoc test for mixed methods analysis. From baseline to timepoint 3 (T3, 4 months) daily energy intake was consistent with cohort estimated energy requirements (EER). However, consumption ranged from 37% to 167% of EER, indicating a large variation of intakes. Energy consumed from ULP ranged from 30% to 34% (p = 0.6875). Baseline tryptophan intake (x¯ = 1139mg) was within the suggested target, whilst fibre intake (x¯ = 23g) was below the recommended intake. Neither saw significant changes from baseline to T3. Fibre intake was positively correlated with mental and physical well-being at baseline (x¯ = 23.1g, M: r = 0.474, p = 0.04, P: r = 0.608, p = 0.007), and timepoint 2 (x¯ = 31.5g, M: r = 0.647,p = 0.026, P: r = 0.780, p = 0.004) but not at T3. In addition, baseline consumption of sugar (x¯ = 18g) and poly-unsaturated fats (x¯ = 15g) were both negatively correlated with mental and physical well-being. Overall, no significant dietary changes were evident from baseline to mid-year collection in a first-year medical intern cohort during hospital placements. Fibre was significantly associated with mental and physical well-being, building on current understanding of fibre’s role in the MGBA. Planned metabolite analysis will explore the mechanisms of proposed microbiome-accessible nutrients alongside diet, well-being, and microbiota data. Findings from this study will identify how diet-microbiome interactions change under stress, with wider positive implications on intense workplace environments with the aim to preserve individual wellbeing.

Nomogram model for predicting cancer-specific mortality in patients with early-onset colorectal cancer: a competing risk analysis insight from the SEER database and an external validation cohort.

Early-onset colorectal cancer (EOCRC) is increasing in incidence and poses a growing threat. Urgent research is needed, especially in survival analysis, to enhance comprehension and treatment strategies. This study aimed to explore the risk factors associated with cancer-specific mortality (CSM) and other-cause mortality (OCM) in patients with EOCRC. Additionally, the study aimed to develop a nomogram predicting CSM using a competitive risk model and validate its accuracy through the use of training, using internal and external cohorts. Data from EOCRC patients were collected from the Surveillance, Epidemiology, and End Results (SEER) database (2008-2017). EOCRC patients who were treated at a tertiary hospital in northeast China between 2014 and 2020 were also included in the study. The SEER data were divided into the training and validation sets at a 7:3 ratio. A univariate Cox regression model was employed to identify prognostic factors. Subsequently, multivariate Cox regression models were applied to ascertain the presence of independent risk factors. A nomogram was generated to visualize the results, which were evaluated using the concordance index (C-index), area under the curve (AUC), and calibration curves. The clinical utility was assessed via decision curve analysis (DCA). Multivariable Cox regression analysis demonstrated that factors such as race, tumor differentiation, levels of carcinoembryonic antigen (CEA), marital status, histological type, American Joint Committee on Cancer (AJCC) stage, and surgical status were independent risk factors for CSM in EOCRC patients. In addition, age, gender, chemotherapy details, CEA levels, marital status, and AJCC stage were established as independent risk factors for OCM in individuals diagnosed with EOCRC. A nomogram was developed using the identified independent risk factors, demonstrating excellent performance with a C-index of 0.806, 0.801, and 0.810 for the training, internal validation, and external validation cohorts, respectively. The calibration curves and AUC further confirmed the accuracy and discriminative ability of the nomogram. Furthermore, the DCA results indicated that the model had good clinical value. In this study, a competing risk model for CSM was developed in EOCRC patients. The model demonstrates a high level of predictive accuracy, providing valuable insights into the treatment decision-making process.

An androgen receptor-based signature to predict prognosis and identification of ORC1 as a therapeutical target for prostate adenocarcinoma.

Aberrant activation of androgen receptor (AR) signaling plays a crucial role in the progression of prostate adenocarcinoma (PRAD) and contributes significantly to the development of enzalutamide resistance. In this study, we aimed to identify a novel AR-driven signature that can predict prognosis and endows potentially reveal novel therapeutic targets for PRAD. The Seurat package was used to preprocess the single-cell RNA sequencing (scRNA-seq). Differentially expressed genes were visualized using limma and pheamap packages. LASSO and multi-variate Cox regression models were established using glmnet package. The package "Consensus Cluster Plus" was utilized to perform the consensus clustering analysis. The biological roles of origin recognition complex subunit 1 (ORC1) in PRAD were determined by gain- and loss-of-function studies in vitro and in vivo. We characterized the scRNA-seq data from GSE99795 and identified 10 AR-associated genes (ARGs). The ARGs model was trained and validated in internal and external cohorts. The ARGs were identified as an independent hazard factor in PRAD and correlated with clinical risk characteristics. In addition, the ARGs were found to be correlated with somatic tumor mutation burden (TMB) levels. Two groups that have distinct prognostic and molecular features were identified through consensus clustering analysis. ORC1 was identified as a critical target among these ARGs, and it ORC1 promoted proliferation and stem-like properties of PRAD cells. Chromatin immunoprecipitation (ChIP)-qPCR assay confirmed that AR could directly bind the promoter of ORC1. Activated AR/ORC1 axis contributed to enzalutamide resistance, and targeting ORC1 rendered PRAD cells more susceptible to enzalutamide. This study defines an AR-driven signature that AR activates ORC1 expressions to promote PRAD progression and enzalutamide resistance, which may provide novel targets for PRAD treatment.

Artificial Intelligence Predicts Hospitalization for Acute Heart Failure Exacerbation in Patients Undergoing Myocardial Perfusion Imaging.

Heart failure (HF) is a leading cause of morbidity and mortality in the United States and worldwide, with a high associated economic burden. This study aimed to assess whether artificial intelligence models incorporating clinical, stress test, and imaging parameters could predict hospitalization for acute HF exacerbation in patients undergoing SPECT/CT myocardial perfusion imaging. Methods: The HF risk prediction model was developed using data from 4,766 patients who underwent SPECT/CT at a single center (internal cohort). The algorithm used clinical risk factors, stress variables, SPECT imaging parameters, and fully automated deep learning-generated calcium scores from attenuation CT scans. The model was trained and validated using repeated hold-out (10-fold cross-validation). External validation was conducted on a separate cohort of 2,912 patients. During a median follow-up of 1.9 y, 297 patients (6%) in the internal cohort were admitted for HF exacerbation. Results: The final model demonstrated a higher area under the receiver-operating-characteristic curve (0.87 ± 0.03) for predicting HF admissions than did stress left ventricular ejection fraction (0.73 ± 0.05, P < 0.0001) or a model developed using only clinical parameters (0.81 ± 0.04, P < 0.0001). These findings were confirmed in the external validation cohort (area under the receiver-operating-characteristic curve: 0.80 ± 0.04 for final model, 0.70 ± 0.06 for stress left ventricular ejection fraction, 0.72 ± 0.05 for clinical model; P < 0.001 for all). Conclusion: Integrating SPECT myocardial perfusion imaging into an artificial intelligence-based risk assessment algorithm improves the prediction of HF hospitalization. The proposed method could enable early interventions to prevent HF hospitalizations, leading to improved patient care and better outcomes.

Radiomics model based on MRI to differentiate spinal multiple myeloma from metastases: A two-center study

PurposeSpinal multiple myeloma (MM) and metastases are two common cancer types with similar imaging characteristics, for which differential diagnosis is needed to ensure precision therapy. The aim of this study is to establish radiomics models for effective differentiation between them. MethodsEnrolled in this study were 263 patients from two medical institutions, including 127 with spinal MM and 136 with spinal metastases. Of them, 210 patients from institution I were used as the internal training cohort and 53 patients from Institution II were used as the external validation cohort. Contrast-enhanced T1-weighted imaging (CET1) and T2-weighted imaging (T2WI) sequences were collected and reviewed. Based on the 1037 radiomics features extracted from both CET1 and T2WI images, Logistic Regression (LR), AdaBoost (AB), Support Vector Machines (SVM), Random Forest (RF), and multiple kernel learning based SVM (MKL-SVM) were constructed. Hyper-parameters were tuned by five-fold cross-validation. The diagnostic efficiency among different radiomics models was compared by accuracy (ACC), sensitivity (SEN), specificity (SPE), area under the ROC curve (AUC), YI, positive predictive value (PPV), negative predictive value (NPY), and F1-score. ResultsBased on single-sequence, the RF model outperformed all other models. All models based on T2WI images performed better than those based on CET1. The efficiency of all models was boosted by incorporating CET1 and T2WI sequences, and the MKL-SVM model achieved the best performance with ACC, AUC, and F1-score of 0.862, 0.870, and 0.874, respectively. ConclusionsThe radiomics models constructed based on MRI achieved satisfactory diagnostic performance for differentiation of spinal MM and metastases, demonstrating broad application prospects for individualized diagnosis and treatment.

Predicting rectal cancer tumor budding grading based on MRI and CT with multimodal deep transfer learning: A dual-center study

ObjectiveTo investigate the effectiveness of a multimodal deep learning model in predicting tumor budding (TB) grading in rectal cancer (RC) patients. Materials and methodsA retrospective analysis was conducted on 355 patients with rectal adenocarcinoma from two different hospitals. Among them, 289 patients from our institution were randomly divided into an internal training cohort (n = 202) and an internal validation cohort (n = 87) in a 7:3 ratio, while an additional 66 patients from another hospital constituted an external validation cohort. Various deep learning models were constructed and compared for their performance using T1CE and CT-enhanced images, and the optimal models were selected for the creation of a multimodal fusion model. Based on single and multiple factor logistic regression, clinical N staging and fecal occult blood were identified as independent risk factors and used to construct the clinical model. A decision-level fusion was employed to integrate these two models to create an ensemble model. The predictive performance of each model was evaluated using the area under the curve (AUC), DeLong's test, calibration curve, and decision curve analysis (DCA). Model visualization Gradient-weighted Class Activation Mapping (Grad-CAM) was performed for model interpretation. ResultsThe multimodal fusion model demonstrated superior performance compared to single-modal models, with AUC values of 0.869 (95% CI: 0.761–0.976) for the internal validation cohort and 0.848 (95% CI: 0.721–0.975) for the external validation cohort. N-stage and fecal occult blood were identified as clinically independent risk factors through single and multivariable logistic regression analysis. The final ensemble model exhibited the best performance, with AUC values of 0.898 (95% CI: 0.820–0.975) for the internal validation cohort and 0.868 (95% CI: 0.768–0.968) for the external validation cohort. ConclusionMultimodal deep learning models can effectively and non-invasively provide individualized predictions for TB grading in RC patients, offering valuable guidance for treatment selection and prognosis assessment.

Psychometric Properties of Entrustable Professional Activity-Based Objective Structured Clinical Examinations During Transition From Undergraduate to Graduate Medical Education: A Generalizability Study.

The objective structured clinical examination (OSCE) assesses clinical competence in health sciences education. There is little research regarding the reliability and validity of using an OSCE during the transition from undergraduate to graduate medical education. The goal of this study was to measure the reliability of a unique 2-rater entrustable professional activity (EPA)-based OSCE format for transition to internship using generalizability theory for estimating reliability. During the 2018 to 2022 academic years, 5 cohorts of interns (n = 230) at the University of Iowa Hospital and Clinics participated in a 6-station OSCE assessment delivered during orientation. A univariate and multivariate generalizability study (G study) was conducted on the scores generated by the 3 cases in the orientation OSCE that shared the 2-rater format. This analysis was supplemented with an associated decision study (D study). The univariate G study for the cases that used a simulated patient and a faculty rater demonstrated that this OSCE generated a moderately reliable score with 3 cases. The D study showed that increasing the OCSE to 12 cases yielded a mean score reliable enough (G = 0.76) for making high-stakes normative decisions regarding remediation and readiness to practice. The universe score correlation between 2 types of raters was 0.398. The faculty ratings displayed a larger proportion of universe (true) score variance and yielded a more reliable (G = 0.433) score compared with the standardized patient ratings (G = 0.337). This study provides insight into the development of an EPA-based OSCE. The univariate G study demonstrated that when using the 2 rater types, this assessment could generate a moderately reliable score with 3 cases. The multivariate G study showed that the 2 types of raters assessed different aspects of clinical skills and faculty raters were more reliable.

Predicting three or more metastatic nodes using contrast-enhanced lymphatic US findings in early breast cancer

ObjectivesTo develop and validate a nomogram for predicting ≥ 3 metastatic axillary lymph nodes (ALNs) in early breast cancer with no palpable axillary adenopathy by clinicopathologic data, contrast-enhanced (CE) lymphatic ultrasound (US), and grayscale findings of sentinel lymph nodes (SLNs).Materials and methodsWomen with T1-2N0 invasive breast cancer were consecutively recruited for the CE lymphatic US. Patients from Center 1 were grouped into development and internal validation cohorts at a ratio of 2:1. The external validation cohort was constructed from Center 2. The clinicopathologic data and US findings of SLNs were analyzed. A nomogram was developed to predict women with ≥ 3 metastatic ALNs. Nomogram performance was assessed with the area under the receiver operating characteristic curve (AUC) and calibration curve analysis.ResultsOne hundred seventy-nine from Center 1 were considered the development cohorts. The remaining 90 participants from Center 1 were internal cohorts and 197 participants from Center 2 were external validation cohorts. The US findings of no enhancement (odds ratio (OR), 15.3; p = 0.01), diffuse (OR, 19.1; p = 0.01) or focal eccentric (OR, 27.7; p = 0.003) cortical thickening, and absent hilum (OR, 169.7; p < 0.001) were independently associated with ≥ 3 metastatic ALNs. Compared to grayscale US or CE lymphatic US alone, the nomogram showed the highest AUC of 0.88 (0.85, 0.91). The nomogram showed a calibration slope of 1.0 (p = 0.80–0.81; Brier = 0.066–0.067) in validation cohorts in predicting ≥ 3 metastatic ALNs.ConclusionPatients likely to have ≥ 3 metastatic ALNs were identified by combining the lymphatic and grayscale US findings of SLNs. Our nomogram could aid in multidisciplinary treatment decision-making.Trial registrationThis trial is registered on www.chictr.org.cn: ChiCTR2000031231. Registered March 25, 2020.Critical relevance statementA nomogram combining lymphatic CEUS and grayscale US findings of SLNs could identify early breast cancer patients with low or high axillary tumor burden preoperatively, which is more applicable to the Z0011 era. Our nomogram could be useful in aiding multidisciplinary treatment decision-making for patients with early breast cancer.Key points• CEUS can help identify and diagnose SLN in early breast cancer preoperatively.• Combining lymphatic and grayscale US findings can predict axillary tumor burden.• The nomogram showed a high diagnostic value in validation cohorts.Graphical

Identification and validation of sepsis subphenotypes using time-series data

PurposeThe recognition of sepsis as a heterogeneous syndrome necessitates identifying distinct subphenotypes to select targeted treatment. MethodsPatients with sepsis from the MIMIC-IV database (2008–2019) were randomly divided into a development cohort (80%) and an internal validation cohort (20%). Patients with sepsis from the ICU database of Peking University People's Hospital (2008–2022) were included in the external validation cohort. Time-series k-means clustering analysis and dynamic time warping was performed to develop and validate sepsis subphenotypes by analyzing the trends of 21 vital signs and laboratory indicators within 24 h after sepsis onset. Inflammatory biomarkers were compared in the ICU database of Peking University People's Hospital, whereas treatment heterogeneity was compared in the MIMIC-IV database. FindingsThree sub-phenotypes were identified in the development cohort. Type A patients (N = 2525, 47%) exhibited stable vital signs and fair organ function, type B (N = 1552, 29%) was exhibited an obvious inflammatory response and stable organ function, and type C (N = 1251, 24%) exhibited severely impaired organ function with a deteriorating tendency. Type C demonstrated the highest mortality rate (33%) and levels of inflammatory biomarkers, followed by type B (24%), whereas type A exhibited the lowest mortality rate (11%) and levels of inflammatory biomarkers. These subphenotypes were confirmed in both the internal and external cohorts, demonstrating similar features and comparable mortality rates. In type C patients, survivors had significantly lower fluid intake within 24 h after sepsis onset (median 2891 mL, interquartile range (IQR) 1530–5470 mL) than that in non-survivors (median 4342 mL, IQR 2189–7305 mL). For types B and C, survivors showed a higher proportion of indwelling central venous catheters (p < 0.05). ConclusionThree novel phenotypes of patients with sepsis were identified and validated using time-series data, revealing significant heterogeneity in inflammatory biomarkers, treatments, and consistency across cohorts.

Assessing brain involvement in Fabry disease with deep learning and the brain-age paradigm.

While neurological manifestations are core features of Fabry disease (FD), quantitative neuroimaging biomarkers allowing to measure brain involvement are lacking. We used deep learning and the brain-age paradigm to assess whether FD patients' brains appear older than normal and to validate brain-predicted age difference (brain-PAD) as a possible disease severity biomarker. MRI scans of FD patients and healthy controls (HCs) from a single Institution were, retrospectively, studied. The Fabry stabilization index (FASTEX) was recorded as a measure of disease severity. Using minimally preprocessed 3D T1-weighted brain scans of healthy subjects from eight publicly available sources (N = 2160; mean age = 33 years [range 4-86]), we trained a model predicting chronological age based on a DenseNet architecture and used it to generate brain-age predictions in the internal cohort. Within a linear modeling framework, brain-PAD was tested for age/sex-adjusted associations with diagnostic group (FD vs. HC), FASTEX score, and both global and voxel-level neuroimaging measures. We studied 52 FD patients (40.6 ± 12.6 years; 28F) and 58 HC (38.4 ± 13.4 years; 28F). The brain-age model achieved accurate out-of-sample performance (mean absolute error = 4.01 years, R2 = .90). FD patients had significantly higher brain-PAD than HC (estimated marginal means: 3.1 vs. -0.1, p = .01). Brain-PAD was associated with FASTEX score (B = 0.10, p = .02), brain parenchymal fraction (B = -153.50, p = .001), white matter hyperintensities load (B = 0.85, p = .01), and tissue volume reduction throughout the brain. We demonstrated that FD patients' brains appear older than normal. Brain-PAD correlates with FD-related multi-organ damage and is influenced by both global brain volume and white matter hyperintensities, offering a comprehensive biomarker of (neurological) disease severity.

Prognostic and immunotherapeutic potential of regulatory Tcell-associated signature in ovarian cancer.

Tumour-induced immunosuppressive microenvironments facilitate oncogenesis, with regulatory T cells (Tregs) serving as a crucial component. The significance of Treg-associated genes within the context of ovarian cancer (OC) remains elucidated insufficiently. Utilizing single-cell RNA sequencing (scRNA-Seq) for the identification of Treg-specific biomarkers, this investigation employed single-sample gene set enrichment analysis (ssGSEA) for the derivation of a Treg signature score. Weighted gene co-expression network analysis (WGCNA) facilitated the identification of Treg-correlated genes. Machine learning algorithms were employed to determine an optimal prognostic model, subsequently exploring disparities across risk strata in terms of survival outcomes, immunological infiltration, pathway activation and responsiveness to immunotherapy. Through WGCNA, a cohort of 365 Treg-associated genes was discerned, with 70 implicated in the prognostication of OC. A Tregs-associated signature (TAS), synthesized from random survival forest (RSF) and Least Absolute Shrinkage and Selection Operator (LASSO) algorithms, exhibited robust predictive validity across both internal and external cohorts. Low TAS OC patients demonstrated superior survival outcomes, augmented by increased immunological cell infiltration, upregulated immune checkpoint expression, distinct pathway enrichment and differential response to immunotherapeutic interventions. The devised TAS proficiently prognosticates patient outcomes and delineates the immunological milieu within OC, offering a strategic instrument for the clinical stratification and selection of patients.

Abstract 2997: Identification of novel long non-coding RNAs in mantle cell lymphoma patients

Abstract Mantle Cell Lymphoma (MCL) is a rare and incurable type of B-cell non-Hodgkin lymphoma that affects about 4,000 people per year in the United States. Despite intense research to identify novel therapeutic approaches, the median progression-free survival after first-line treatment is four years. For this reason, the discovery of novel therapeutic targets remains crucial. Over half of the human genome is actively transcribed as non-coding RNAs. Among them, the non-coding RNAs that are longer than 200 nucleotides are named long non-coding RNAs (lncRNAs). lncRNAs have been shown to play essential roles in several human diseases, including cancer, by modulating cell cycle, immune response, or stemness through their interactions with DNA, protein, and RNA. For this reason, we decided to perform an RNA sequencing (RNA-Seq) profiling to identify the dysregulated lncRNAs in MCL. In more detail, we performed RNA-Seq in a cohort of 18 MCL patients compared to normal B-cells from the peripheral blood of healthy donors (HD). Reads were first trimmed for quality filtering, and adapters were removed using Trim Galore. Trimmed reads were mapped to the human genome (HG38 assembly) using HISAT 2. All the reads mapped inside the genomic coordinates of lncRNAs, reported in the GTF annotation file of GENCODE v43, were counted by featureCounts. Differential expression analysis between MCL samples vs HD was performed by LIMMA. To validate the results obtained from our internal MCL cohort, we also downloaded publicly available RNA-Seq data of 16 MCL patients and analyzed them using the previously described pipeline. Our analyses revealed several dysregulations in lncRNA expression occurring in MCL. Precisely, we identified 385 upregulated and 387 downregulated lncRNAs (|log2FC| &amp;gt; 0.58 and adjusted p-value &amp;lt; 0.05) in our internal MCL cohort, and 63% of them were also confirmed in the external cohort. Notably, the lncRNA called MALAT1 showed the highest significant upregulation in both cohorts, reinforcing this lncRNA's pivotal role in cancer development. To study the potential involvement of MALAT1 in MCL, we compared the MCL samples with higher MALAT1 expression (upper quartile) against those with lower expression (lower quartile) in both cohorts. The identified differentially expressed genes were used as input for a mechanistic network-based topological pathway analysis using MITHrIL. The analysis revealed that the KEGG pathway "Cell Cycle" was among the top upregulated pathways in MCL samples with high MALAT1 expression, suggesting the relevance of this lncRNA in promoting cellular proliferation, as also reported in the literature for other tumor types. In conclusion, our is the first study to describe the novel dysregulation signature of lncRNAs in MCL in two independent cohorts. Citation Format: Alessandro LaFerlita, Satishkumar Singh, Anuvrat Sircar, Robert Baiocchi, Natarajan Muthusamy, Lapo Alinari, Narendranath Epperla, Lalit Sehgal. Identification of novel long non-coding RNAs in mantle cell lymphoma patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2997.

Abstract 6826: Comparative analysis of a predictive transcriptomic model and functional gene expression signatures (FGES) for tertiary lymphoid structure (TLS) detection in lung adenocarcinoma (LUAD)

Abstract Background: TLS are recognized as significant biomarkers for patient prognosis and therapy selection. In current clinical practice, TLS are often detected by manual assessment of H&amp;E slides, which has been associated with reproducibility issues. Here, we developed a transcriptomic model using TLS density to stratify LUAD patients into TLS-high and -low groups and compared it with 5 published TLS FGES to assess the predictive value for patient outcomes. Design: A regression model to predict TLS density (units(u)/mm²) was created using RNA expression data and histological mapping of 415 TCGA-LUAD samples. Features were chosen based on differential gene expression analysis of sample groups with low (&amp;lt;Q1, 0.26 u/mm²) and high (&amp;gt;Q3, 2.04 u/mm²) TLS density. The model was trained on 269 samples and independently validated on 146 TCGA and 9 internal samples. The RNA model and five previously reported TLS FGES were compared by Wilcoxon paired test to Spearman’s ρ and F1-scores from bootstrapped validation set. The resulting predictions stratified immunotherapy-treated samples from SU2C-MARK (n=58) and GSE182328 (n=28) into TLS-high and TLS-low groups by Q2 threshold for TLS density (0.78 u/mm²) for further univariate (Kaplan-Meier) and multivariate (Cox model, adjusted for age and sex) analyses for overall survival (OS) with log-rank tests. P-values were adjusted with Bonferroni correction, and 95% confidence intervals (CIs) were provided. Results: Statistically significant upregulated genes in samples with high TLS density included FDCSP, MS4A1, and CXCL13 (logFC ≥ 0.95, FDR&amp;lt;0.01). The transcriptomic model showed confident superiority in correlation with TLS density compared to the best FGES in TCGA samples (ρ=0.53±0.03 vs 0.41±0.03, p&amp;lt;0.001). The TLS model also outperformed the highest performing FGES when stratifying TCGA samples into TLS-high and TLS-low groups (F1=0.73±0.02 vs 0.69±0.02, p&amp;lt;0.005). In the internal cohort, the model demonstrated robust performance (F1=0.76±0.05, ρ=0.81±0.04) and outperformed all FGES except one (FGES F1=0.78±0.05, ρ=0.86±0.05). The TLS-high group defined by the TLS RNA model had a significantly better prognosis on immunotherapy cohorts (p=0.01; 1 year OS 49% for TLS-low vs 70% for TLS-high), whereas no discernible difference in OS was identified for the 5 TLS FGES. The TLS model had a log hazard ratio (logHR) of -3.11 (CI:-5.31,-0.91; p&amp;lt;0.05, Harrell's C-index=0.64), indicative of a stronger association with OS than of the TLS FGES (logHR=-2.11; CI:-4.1,-0.08; p=ns; C-index=0.59). Conclusions: Comparative analysis of the proposed transcriptomic model for TLS detection in LUAD samples demonstrated improved performance metrics and prognostic value compared to previously reported TLS FGES, affirming its potential for guiding therapeutic decision-making. Citation Format: Nadezhda Lukashevich, Daniil Dymov, Aleksandr Sarachakov, Sofya Kust, Anna Love, Ivan Valiev, Dmitry Ivchenkov, Alexander Bagaev, Viktor Svekolkin, Nikita Kotlov, Vladimir Kushnarev. Comparative analysis of a predictive transcriptomic model and functional gene expression signatures (FGES) for tertiary lymphoid structure (TLS) detection in lung adenocarcinoma (LUAD) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6826.