ROC AUC Research Articles

Machine Learning-Based Analysis Reveals Triterpene Saponins and Their Aglycones in Cimicifuga racemosa as Critical Mediators of AMPK Activation.

Cimicifuga racemosa (CR) extracts contain diverse constituents such as saponins. These saponins, which act as a defense against herbivores and pathogens also show promise in treating human conditions such as heart failure, pain, hypercholesterolemia, cancer, and inflammation. Some of these effects are mediated by activating AMP-dependent protein kinase (AMPK). Therefore, comprehensive screening for activating constituents in a CR extract is highly desirable. Employing machine learning (ML) techniques such as Deep Neural Networks (DNN), Logistic Regression Classification (LRC), and Random Forest Classification (RFC) with molecular fingerprint MACCS descriptors, 95 CR constituents were classified. Calibration involved 50 randomly chosen positive and negative controls. LRC achieved the highest overall test accuracy (90.2%), but DNN and RFC surpassed it in precision, sensitivity, specificity, and ROC AUC. All CR constituents were predicted as activators, except for three non-triterpene compounds. The validity of these classifications was supported by good calibration, with misclassifications ranging from 3% to 17% across the various models. High sensitivity (84.5-87.2%) and specificity (84.1-91.4%) suggest suitability for screening. The results demonstrate the potential of triterpene saponins and aglycones in activating AMP-dependent protein kinase (AMPK), providing the rationale for further clinical exploration of CR extracts in metabolic pathway-related conditions.

Comparison of Machine Learning Tree-Based Algorithms to Predict Future Paratuberculosis ELISA Results Using Repeat Milk Tests.

Machine learning algorithms have been applied to various animal husbandry and veterinary-related problems; however, its use in Johne's disease diagnosis and control is still in its infancy. The following proof-of-concept study explores the application of tree-based (decision trees and random forest) algorithms to analyze repeat milk testing data from 1197 Canadian dairy cows and the algorithms' ability to predict future Johne's test results. The random forest models using milk component testing results alongside past Johne's results demonstrated a good predictive performance for a future Johne's ELISA result with a dichotomous outcome (positive vs. negative). The final random forest model yielded a kappa of 0.626, a roc AUC of 0.915, a sensitivity of 72%, and a specificity of 98%. The positive predictive and negative predictive values were 0.81 and 0.97, respectively. The decision tree models provided an interpretable alternative to the random forest algorithms with a slight decrease in model sensitivity. The results of this research suggest a promising avenue for future targeted Johne's testing schemes. Further research is needed to validate these techniques in real-world settings and explore their incorporation in prevention and control programs.

Integrating IoT and Artificial intelligent for Enhanced Electric Vehicle Charging and Autonomous Driving for sustainable transportation

This research explores the confluence of Internet of things and artificial intelligence towards fostering electric vehicle charging and autonomous driving for sustainable transportation. By meticulously collecting and preprocessing, and executing multiple machine learning models such as Artificial Neural Network , Decision Trees , Naive Bayes , and Random Forest , the research delves into charging infrastructure and easy autonomous driving. Particularly, real-time datasets, such as EV station data, road geography and mapping data, weather and telematics data, were properly collected and processed . Then, the preprocessing stage, which included data cleaning, normalization, and feature engineering had improved the datasets’ predictability. Consequently, the datasets such as charging recommendation, predicting energy use, predicting of optimal route, had produced highly accurate training and testing for forecasting elelctric vehicles’ demands. The ML models indicated their proficiency in predicting not only the optimal charging schedules, but also predictive energy consumption, and predictive routes. Specifically, ANN with 0.945 held the highest precision, while DT 0.912, NB 0.887, and RF 0.819 were almost identical . Likewise, the recall scores, F1 scores, and AUC ROC scores also coincided the prized outcome of the models. The findings of the research are groundbreaking for both transportation sector, policymakers, as well as urban planners and stakeholders at large. Following the data-fed recommendations would not be excessively challenging, and they would yield favourable results whether the stakeholders are dealing with EV charging infrastructure, or dealing with traffic congestions, or are establishing their own sustainable solution for future mobility via not just electric vehicles but also other tools and devices. Hence, we are hopeful that our research serves as a lighthouse for the transportation domain towards a greener future.

Together is better – RibScore and SCARF in the prediction of pulmonary complications and association with SSRF

BackgroundOptimal treatment of patients with rib fractures requires identification of those patients at risk of pulmonary complications. It is also important to determine which patients would benefit from Surgical Stabilisation of Rib Fractures (SSRF). This study aims to validate two scoring systems (RibScore and SCARF score) in predicting complications and association with SSRF in an Australian trauma population. Clinical observation suggests that complications and criteria for SSRF is associated with anatomical and physiological factors. Therefore it is hypothesized that utilisation of an anatomical (RibScore) and physiological (SCARF) in conjunction will have improved predictive ability. MethodRetrospective cohort study of rib fracture patients admitted to an Australian Level I trauma centre from Jan 2017 to Jan 2021. RibScore and SCARF score were calculated. Multivariate logistic regression was performed to determine risk factors associated with complications and SSRF, as well the scoring systems’ ability via ROC AUC. Results1157 patients were included. Higher median RibScore (1vs0; p < 0.001) and SCARF score (3vs1, p < 0.001) was associated with development of complications. Similarly for SSRF, RibScore (3vs0; p < 0.001), SCARF score (3vs1; p < 0.001) were higher. On multivariate analysis, increasing RibScore and SCARF score were associated with an increased risk of respiratory failure, pneumonia, death, and SSRF. The sensitivity for a patient with a high risk score in either RibScore or SCARF increased to 96.3 % in identifying pulmonary complications (from 66.7 % in RibScore and 88 % in SCARF, when used individually) and 91.9 % in identifying association with SSRF (from 86.5 % in RibScore and 70.3 % in SCARF). ConclusionRibScore and SCARF score demonstrate predictive ability for complications and SSRF in an Australian trauma rib fracture population. Combining a radiological score with a clinical scoring system demonstrates improved sensitivity over each score individually for identifying patients at risk of complications from rib fractures, those who may require SSRF, and those who are low risk. Study typeRetrospective Cohort Study Level of evidenceLevel III

Is there a femoral morphology at risk of a cervical or pertrochanteric fracture? Description and validation of the “Neck shaft ratio” and the “Intertrochanteric distal ratio” after analyzing a continuous cohort of 126 bilateral fractures of the proximal femur

BackgroundProximal femur fractures constitute a public health concern given their high frequency and the aging population. The frequency of a contralateral fracture occurring can reach up to 15% of cases. Certain historical or demographic factors constitute risk factors for refracture, but the type of fracture, either of the femoral neck (FN) or pertrochanteric (PT), cannot be predicted. The objectives of this retrospective study were: firstly, to analyze several anatomical markers in order to determine whether they predispose to a certain type of fracture in the event of contralateral refracture, and secondly, to determine the predictive power of these possible radiological markers for the type of contralateral fracture. HypothesisThe hypothesis was that the “Neck shaft ratio” (NSR) and the “Intertrochanteric distal ratio” (ITDR) made it possible to determine a proximal femoral morphology at risk of FN and/or PT fracture in the event of a second fracture. Material and methodsThis continuous retrospective single-center series from January 2011 to December 2019 of patients who presented with bilateral fractures of the proximal femurs was analyzed. Radiographs, taken after the first fracture, of the contralateral femur were studied. Morphological measurements previously described in the literature were carried out as well as the NSR (ratio of the medial cortical thicknesses of the femoral neck at its narrowest and basicervical level) and the ITDR (ratio of the medial diaphyseal cortical thicknesses at 5mm and 20mm from the distal portion of the lesser trochanter). One hundred and twenty-six patients were included: 46/126 (36.5%) had bilateral FN, 50/126 (39.7%) bilateral PT and 30/126 (23.8%) one of each. ResultsOnly NSR and ITDR were significant predictive measures for FN or PT fracture type [0.54±0.11 vs. 0.81±0.16 (p<0.0001) and 0. 85±0.1 vs. 0.68±0.1 (p<0.0001), respectively]. These two ratios had an excellent predictive value for the type of fracture occurring on the contralateral side [NSR: AUC ROC = 0.91 (95% CI; 0.86–0.96); ITDR: AUC ROC = 0.81 (95% CI; %, 0.74–0.89)]. The NSR had excellent intra- and inter-observer reproducibility with an interclass correlation coefficient of 0.93 [95% CI: 0.86–0.97] and 0.91 [95% CI: 0.82–0.96] respectively, the same also applied for the ITDR with interclass correlation coefficient values of 0.93 [95% CI: 0.87–0.97] and 0.86 [95% CI: 0.73–0.93] respectively. DiscussionThe NSR and ITDR ratios offer a simple and reproducible means to predict a morphological predisposition to a certain fracture type, respectively an FN and PT on the side contralateral to an initial osteoporotic proximal femur fracture. A prospective cohort study would be useful in defining a possible prognostic nature on the occurrence and/or time until refracture. Level of evidenceIII; retrospective control case.

Firm failure prediction using genetic programming generated features

Many studies on predicting firm failure have focused on finding new features that improve the accuracy of the models. In this paper, genetic programming (GP) is used for this purpose. The main problem in GP is to specify a function that evaluates the fitness of the feature. Direct optimization of a machine learning (ML) model that uses a generated feature in most cases leads to high computational costs since evolving a population of N programs over G generations while evaluating each model using K-fold cross-validation requires N*G*K model learning cycles. Thus, many researchers use scores that measure the relationship of the generated features to the class label. However, our empirical analysis shows that most such scores correlate poorly with ML model performance. The novelty of our work is that we introduce several ways of combining different scores into a single measure of expected model performance. Experimental results on data from Hungarian firms (7167 observations, class imbalance 9.37) using five ML models (Logistic Regression, Random Forest, Gradient Boosting, Histogram Boosting, and AdaBoost) prove that the proposed way of setting the fitness function increases the ROC AUC of the listed models by 6.6%, 5.2%, 6.8%, 5.5% and 5.2% respectively. Moreover, by applying the found formula to the data from Czech firms (3872 observations, class imbalance of 74.92), which were not used for the feature search, we obtained increases in ROC AUC by 13.1%, 11.8%, 14.9%, 9.9%, and 8.2%, respectively. This indicates that the proposed method allows to find universal features, which opens the way to build effective models in case of insufficient data (small number of observations, extreme imbalance, etc.).

Association between Physical Activity and Menstrual Cycle Disorders in Young Athletes.

Our study aims to evaluate clinical predictors of menstrual cycle disorders in female athletes who compete in running disciplines. This is a prospective observational study. Women were recruited between January and May 2022. Fifty-three patients were enrolled and completed a questionnaire about menstrual cycle, physical activity, and food habit characteristics. Of the women in our population, 39.6% had menstrual irregularities and reported a significantly higher number of kilometers run per week (67 vs. 35, p:0.02). The number of kilometers run per week was associated with menstrual irregularities (for 10 km, OR 1.35; IC95% 1.05-1.73; p: 0.02) after adjusting for BMI, age, level of sport and caloric intake. The variable of "km run per week" appeared as a diagnostic indicator of irregular menstrual cycle with statistical significance (AUC ROC curve 0.71, IC95% 0.54-0.86, p-value=0.01) and the cut-off of 65 km run per week is a good indicator of the presence of irregular menstrual cycle (sensitivity (SE) and specificity (SP) of 55% and 81.48%). Menstrual cycle disorders are very frequent in female athletes, and the variable of km run per week may play a role in screening endurance athletes at high risk for these disorders.

Detection of medial vascular calcification in chronic kidney disease based on pulse wave analysis in the frequency domain

Medial vascular calcification (mVC), particularly prevalent in patients with chronic kidney disease (CKD), is a pathological deposition of minerals in the medial layer of the vessel wall and is associated with an increased risk of cardiovascular disease. Currently, there is no cost–effective and non-invasive procedure of mVC assessment in routine clinical practice. We explored whether in–depth analysis of non-invasively recorded peripheral pulse waves can serve as an effective mVC biomarker.The study included 97 CKD patients with histological assessment of mVC in the epigastric artery and pulse wave measurements in the brachial artery. Pulse waves were analysed in the frequency domain to obtain features, which, together with traditional mVC risk factors (i.e., age, sex, and body mass index), were used to build generalized linear models for mVC prediction. The classifiers with the best scores in terms of balanced accuracy were combined in an ensemble establishing the final model.The final, ensembled model, assessed using a leave-one-out cross-validation process, achieved a balanced accuracy equal to 0.87, an accuracy of 0.93, an AUC ROC of 0.91, and an F-score of 0.96. Apart from the features associated with pulse waves, the selected variables included age, sex, body mass index, heart rate and diastolic blood pressure.Analysis of non-invasively recorded peripheral pulse waveforms combined with traditional risk factors, can help to detect mVC in CKD patients and thus, potentially introduce risk-lowering therapeutic strategies at earlier disease stages in the future.

Sensitivity of the African neuropsychology battery memory subtests and learning slopes in discriminating APOE 4 and amyloid pathology in adult individuals in the Democratic Republic of Congo.

The current study examined the sensitivity of two memory subtests and their corresponding learning slope metrics derived from the African Neuropsychology Battery (ANB) to detect amyloid pathology and APOEε4 status in adults from Kinshasa, the Democratic Republic of the Congo. 85 participants were classified for the presence of β-amyloid pathology and based on allelic presence of APOEε4 using Simoa. All participants were screened using CSID and AQ, underwent verbal and visuospatial memory testing from ANB, and provided blood samples for plasma Aβ42, Aβ40, and APOE proteotype. Pearson correlation, linear and logistic regression were conducted to compare amyloid pathology and APOEε4 status with derived learning scores, including initial learning, raw learning score, learning over trials, and learning ratio. Our sample included 35 amyloid positive and 44 amyloid negative individuals as well as 42 without and 39 with APOEε4. All ROC AUC ranges for the prediction of amyloid pathology based on learning scores were low, ranging between 0.56-0.70 (95% CI ranging from 0.44-0.82). The sensitivity of all the scores ranged between 54.3-88.6, with some learning metrics demonstrating good sensitivity. Regarding APOEε4 prediction, all AUC values ranged between 0.60-0.69, with all sensitivity measures ranging between 53.8-89.7. There were minimal differences in the AUC values across learning slope metrics, largely due to the lack of ceiling effects in this sample. This study demonstrates that some ANB memory subtests and learning slope metrics can discriminate those that are normal from those with amyloid pathology and those with and without APOEε4, consistent with findings reported in Western populations.

Evaluation of Approaches for the Assessment of HER2 Expression in Breast Cancer by Radionuclide Imaging Using the Scaffold Protein [99mTc]Tc-ADAPT6.

Due to its small size and high affinity binding, the engineered scaffold protein ADAPT6 is a promising targeting probe for radionuclide imaging of human epidermal growth factor receptor type 2 (HER2). In a Phase I clinical trial, [99mTc]Tc-ADAPT6 demonstrated safety, tolerability and capacity to visualize HER2 expression in primary breast cancer. In this study, we aimed to select the optimal parameters for distinguishing between breast cancers with high and low expression of HER2 using [99mTc]Tc-ADAPT6 in a planned Phase II study. HER2 expression was evaluated in primary tumours and metastatic axillary lymph nodes (mALNs). SPECT/CT imaging of twenty treatment-naive breast cancer patients was performed 2 h after injection of [99mTc]Tc-ADAPT6. The imaging data were compared with the data concerning HER2 expression obtained by immunohistochemical evaluation of samples obtained by core biopsy. Maximum Standard Uptake Values (SUVmax) afforded the best performance for both primary tumours and mALNs (areas under the receiver operating characteristic curve (ROC AUC) of 1.0 and 0.97, respectively). Lesion-to-spleen ratios provided somewhat lower performance. However, the ROC AUCs were still over 0.90 for both primary tumours and mALNs. Thus, lesion-to-spleen ratios should be further evaluated to find if these could be applied to imaging using stand-alone SPECT cameras that do not permit SUV calculations.

Abstract 901: BATCHIE: An active learning platform for scalable combination drug screens

Abstract Large-scale combination drug screens are largely considered intractable due to the immense number of possible combinations. Existing approaches use ad hoc fixed experimental designs then train machine learning models to impute novel combinations. We introduce BATCHIE, an orthogonal approach that adaptively conducts experiments in batches. BATCHIE uses information theory and probabilistic modeling to design each batch to be maximally informative based on the results of previous experiments. BATCHIE is fully modular, allowing any Bayesian probabilistic model to be used and any study constraints to be incorporated while maintaining optimality guarantees. Results: In retrospective simulations on public combination screens, BATCHIE saved 10s of thousands to 100s of thousands of experiments relative to non-adaptive baselines. We conducted a prospective study focusing on pediatric sarcomas with BATCHIE. Our study covered 16 cell lines spanning Ewing sarcoma (EWS), osteosarcoma, rhabdomyosarcoma, as well as non-sarcoma cancers and non-cancer lines. We used a drug library of 206 drugs at two doses. After 15 rounds of BATCHIE-driven data collection, we observed 54K unique combinations, covering 4% of the experimental landscape. On unobserved validation data, the BATCHIE model predictions were highly accurate (Pearson’s rho=0.91, p&amp;lt;10-30) and detected the rare (0.004%) combinations with significant synergy (AUC of ROC=0.85, p&amp;lt;10-5). We further investigated 10 combinations that BATCHIE predicted to have high therapeutic index (TI) for a broad selection of EWS lines, meaning a large differential effect between the predicted viabilities of the control lines and the target lines. We found that the TI scores for the top hits were significantly larger than the rest of the screen (p&amp;lt;10−50), with the median top hit TI score lying in the 98th percentile of observed TI scores. We further validated 6 of the top hits in an ex vivo study on 2 patient-derived EWS samples, again finding significantly large TI values (p&amp;lt;10-13), with the median ex vivo TI score lying in the 96th percentile of observed TI scores. The top hits also exhibited biologically plausible rationales including combining PARP inhibitors with topoisomerase 1 inhibitors and alkylating agents, despite our model utilizing no prior knowledge on the molecular targets of our drug library. Combining PARP inhibitors with topoisomerase 1 inhibitors comprise 3 of the 6 currently open phase II clinical combination trials for EWS. Citation Format: Christopher Tosh, Mauricio Tec, Jessica White, Jeffrey F. Quinn, Glorymar Ibanez Sanchez, Paul Calder, Andrew L. Kung, Filemon S. Dela Cruz, Wesley Tansey. BATCHIE: An active learning platform for scalable combination drug screens [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 901.

Abstract 1845: 11,000-plex aptamer based proteomics of colorectal cancer stools for the discovery of novel disease biomarkers

Abstract Given the morbidity and mortality associated with colorectal cancer, novel biomarkers are clearly warranted, especially for early detection, particularly for advanced adenomas. An earlier 1317-plex aptamer-based proteomic screen of stool proteins revealed 92 proteins to be significantly elevated in CRC stool, with the most discriminatory proteins being MMP9, haptoglobin, myeloperoxidase, fibrinogen, and adiponectin (PMID: 34117903). Results from a recent 2000-plex proteomic screen of CRC stools are detailed below, while a 11,000-plex aptamer based proteomic screen is in progress. Comprehensive proteomic screening revealed 116 proteins to be differentially expressed in CRC stool compared to control stool, with 45 being elevated 2-fold or higher. 37 of these proteins were ELISA validated in 3 independent cohorts. Stool MMP-8, MMP-9, Hemoglobin, PGRP-S, Haptoglobin, Myeloperoxidase and Fibrinogen emerged as being most discriminatory for distinguishing CRC from healthy controls (AUC: 0.91-0.95), while Stool Fibrinogen, MMP-9, Hemoglobin, MMP-8, and PGRP-S best distinguished advanced adenoma from healthy control (HC) stool, with stool Fibrinogen topping the list with an ROC AUC value of 0.86, as detailed in Table 1. Functional pathway analysis revealed a significant over-representation of pathways related to anti-oxidant activity, integrin/receptor binding, cytokines, blood coagulation, and lipoprotein biosynthesis in patients with CRC compared to HC. Nuclear Factor IC (NFIC) and IKZF2 were identified as key regulators of the molecular cascades over-represented in CRC. Stool Fibrinogen, MMP-8, MMP-9, PGRP-S, Haptoglobin, and Myeloperoxidase emerge as promising biomarkers for distinguishing CRC/advanced adenomas/healthy stools, outperforming Hemoglobin. Thus, comprehensive proteomics and independent ELISA validation have uncovered novel stool biomarkers for identifying CRC and advanced adenomas, that warrant further head to head comparison against current diagnostic tests. Citation Format: Chandra Mohan, Kamala Vanarsa, Jessica Castillo, Robert Bresalier, Nicholas Chia. 11,000-plex aptamer based proteomics of colorectal cancer stools for the discovery of novel disease biomarkers [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 1845.

Abstract 7395: Histology-based prognosis prediction using deep learning outperforms and is independent of the MGMT methylation status in patients with glioblastoma

Abstract Introduction: MGMT gene promoter methylation status is today the only biomarker of survival for GBM patients and predicts favorable response to temozolomide (TMZ). Here we propose a deep learning-based model using histology data (H&amp;E slides) to predict the prognosis of patients with Glioblastoma (GBM) treated by standard of care (SOC). Methods: We included TCGA GBM patients and selected only patients treated after 2005 (Stupp protocol start); compliant to WHO 2021 classification; treated by surgery and adjuvant therapy; with a follow-up superior to median overall survival (OS) (14.5 months). Short and long survivor labels were defined based on median OS cut-off. A deep learning pipeline was trained to predict these labels from the H&amp;E slides (baseline surgery samples). First, the slides were automatically segmented into tissue and background, divided into a regular grid of tiles (112 × 112µm), and tiles were vectorized using a vision transformer-based feature extractor trained on H&amp;E slides of 5,500 patients with 16 cancer subtypes (GBM excluded). In the second step, ABMIL deep learning model was trained to predict the survival category using the bag of tile features available from step 1 for each patient. The cohort was divided into five folds with even numbers of short and long survivors. For each fold, 50 ABMIL models were trained on the remaining folds and ensembled to predict the category of the patients in the fold. Training hyperparameters were tuned on a separated cohort excluding GBM patients. Results: After filtering, we analyzed n = 192 TCGA GBM patients including 154 patients with a known MGMT promoter methylation status. Our histology-based model had an accuracy of 0.661 and a ROC AUC of 0.706. It stratified the population into short and long survivors with an Hazard Ratio (HR) of 0.48 (95% CI 0.34-0.69). Our model outperformed the MGMT methylation status that had an HR of 0.57 (95% CI 0.40-0.83). Moreover, our model was independent of MGMT methylation status and the combination of both was able to further improve prognosis prediction: we applied our model independently to the 91 patients with an unmethylated MGMT status (predicted: 43 low, 48 high) and to the 63 patients with a methylated MGMT status (predicted: 30 low, 33 high), and obtained HR of 0.42 (95% CI 0.27-0.67) and 0.49 (95% CI 0.28-0.86) respectively. Conclusion: Having selected patients treated by the latest SOC and classified as GBM by the latest WHO criteria, we propose a model predicting survival based on histological specimens available in routine care, which outperformed the current unique biomarker MGMT methylation status. The combination of this status and our model improves prognosis predictions compared to each strategy alone. Citation Format: Lucas Fidon, Celine Thiriez, Alexandre Grimaldi, Omar Darwiche Domingues, Charles Maussion, Caroline Hoffmann. Histology-based prognosis prediction using deep learning outperforms and is independent of the MGMT methylation status in patients with glioblastoma [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 7395.

Abstract 2295: Circulating T cell receptor repertoire analysis improves cancer early detection

Abstract Background: Cancer screening by liquid biopsy promises to detect cancer early when it can be cured. However, current liquid biopsy approaches based on circulating tumor DNA (ctDNA) detect only a minority of early-stage cancers, because these tumors shed little ctDNA and plasma sample volume is limited. For liquid biopsy to realize its potential for the early detection of cancer, sensitivity for the earliest stage disease must be improved. Methods: We set out to improve sensitivity of liquid biopsy for cancer early detection by exploiting tumor recognition by T cells through sequencing of the circulating T cell receptor repertoire. Using gDNA extracted from blood buffy coats, we studied a cohort of 451 lung cancer patients (86% with stage I disease) and 525 subjects without cancer, enriched for older individuals with a history of smoking. We performed genomic DNA TCR beta chain sequencing to yield a median of 81,088 TCR clonotypes in cancer patients and 112,588 clonotypes in non-cancer controls. We built a TCR sequence similarity nearest neighbor graph to cluster &amp;gt;68M distinct TCR clonotypes into TCR repertoire functional units (RFUs). The TCR frequencies of RFUs were tested for association with cancer status and significantly associated RFUs were combined using a support vector machine model into a TCR RFU cancer score. This model was evaluated by 10-fold cross-validation (CV) and compared to a ctDNA panel of 237 mutation hotspots in 154 lung cancer driver genes (IDT™) and to a panel of 17 lung cancer related protein biomarkers (Olink®) in 87 subjects. Results: We identified 177 cancer-associated TCR RFUs at FDR≤0.10, including 83 RFUs that were enriched in cancer samples with a TCR count log2-fold change between 0.05 and 1.31, and 94 that were enriched in controls with a log2-fold change between 0.07 and 0.40. We observed 56 of the 83 cancer-enriched RFUs had decreasing TCR counts with increasing age across the cohort, coupled with higher counts in cancer patients relative to age-matched controls. The TCR RFU cancer score achieved a cross-validated ROC AUC of 0.69 for cancer status prediction, with the AUC for stage I cancers similar to that of stage II-IV (0.69 vs. 0.70). 44% of stage I cancers (test samples of each CV fold) were detected by the TCR model at a specificity of 80%. We saw a substantial gain in sensitivity for stage I cancer when TCR RFUs were added to ctDNA and proteins, with a ~15-20%-point increase seen at the 90% target specificity level typical for cancer screening tests. Likewise, sensitivity increased at the 99% specificity level used in multi-cancer early detection, with 35% of stage I lung cancers detected, which could not be achieved by any analyte alone. Conclusion: We demonstrate detection of a significant fraction of early lung cancer cases from blood by analyzing the circulating TCR repertoire and show that this signal is complementary to established analytes such as proteins and ctDNA. Citation Format: Yilong Li, Michelle Nahas, Dennis Stephens, Kate Froburg, Emma Hintz, Devin Champagne, Amaneet Lochab, Markus Brown, Jesse Abhyankar, Jasper Braun, Michelle Dinh, María Antonia Fortuño, María-del-Mar Ocon, Andrea Pasquier, Inés María Luque, Christopher W. Seder, Jeffrey A. Borgia, Luis Miguel Seijo, Luis M. Montuenga, Roman Yelensky. Circulating T cell receptor repertoire analysis improves cancer early detection [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 2295.

Advancing Cough Classification: Swin Transformer vs. 2D CNN with STFT and Augmentation Techniques

Coughing, a common symptom associated with various respiratory problems, is a crucial indicator for diagnosing and tracking respiratory diseases. Accurate identification and categorization of cough sounds, specially distinguishing between wet and dry coughs, are essential for understanding underlying health conditions. This research focuses on applying the Swin Transformer for classifying wet and dry coughs using short-time Fourier transform (STFT) representations. We conduct a comprehensive evaluation, including a performance comparison with a 2D convolutional neural network (2D CNN) model, and exploration of two distinct image augmentation methods: time mask augmentation and classical image augmentation techniques. Extensive hyperparameter tuning is performed to optimize the Swin Transformer’s performance, considering input size, patch size, embedding size, number of epochs, optimizer type, and regularization technique. Our results demonstrate the Swin Transformer’s superior accuracy, particularly when trained on classically augmented STFT images with optimized settings (320 × 320 input size, RMS optimizer, 8 × 8 patch size, and an embedding size of 128). The approach achieves remarkable testing accuracy (88.37%) and ROC AUC values (94.88%) on the challenging crowdsourced COUGHVID dataset, marking improvements of approximately 2.5% and 11% increases in testing accuracy and ROC AUC values, respectively, compared to previous studies. These findings underscore the efficacy of Swin Transformer architectures in disease detection and healthcare classification problems.

Abstract 3762: 11,000-plex aptamer based proteomics of bladder cancer urine for the discovery of novel disease biomarkers

Abstract In search of more accurate, non-invasive biomarkers of bladder cancer (BC), a comprehensive aptamer-based screen of urine proteins was conducted using urine samples from a BC cohort, followed by systems biology analyses. Following an initial 1317-plex screen, an extended 11,000 plex aptamer-based proteomic screen is in progress. Based on the initial screen, 30 urine proteins were next ELISA validated in an independent cohort of 68 subjects, and the results are shown in Table 1. Of 21 urine proteins that discriminated BC from urology clinic controls (UC), urine D-dimer displayed the highest ROC AUC value (0.96) and sensitivity (95%) (Table 1). Furthermore, 9 urine proteins significantly discriminated more advanced BC (T1-T4) from earlier stages (Ta-Tis), with IL-8 and MMP-12 being the best performers. Urine MMP12 exhibited the highest specificity of 80% at fixed sensitivity for identifying advanced BC. Systems biology analysis implicated molecular functions related to extracellular matrix, collagen, integrin, heparin and transmembrane tyrosine kinase signaling in BC susceptibility, with HNF4A and NFKB1 emerging as key molecular regulators. STEM analysis of the dysregulated pathways implicated a functional role for the immune system, complement and interleukins in BC disease progression. Urine C2, D-dimer, and Elastase were additionally ELISA validated in a second BC cohort of Chinese ethnicity. Thus, comprehensive proteomics and independent ELISA validation have uncovered novel urine biomarkers for initial BC diagnosis and subsequent disease progression, that warrant further head to head comparison with current diagnostic tests. Citation Format: Chandra Mohan, Kamala Vanarsa, Jessica Castillo, Long Wang, Yair Lotan. 11,000-plex aptamer based proteomics of bladder cancer urine for the discovery of novel disease biomarkers [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 3762.

Deep learning enables early stage prediction of preterm birth using vaginal microbiota

Graphical abstract Abstract Objective Preterm birth (PTB) is one of the leading issues concerning infant health and is a problem that plagues all parts of the world. Vaginal microbial communities have recently garnered attention in the context of PTB; however, the vaginal microbiome varies greatly from individual to individual, and this variation is more pronounced in racially, ethnically, and geographically diverse populations. Additionally, microbial communities have been reported to evolve during the duration of the pregnancy, and capturing such a signature may require higher, more complex modeling paradigms. In this study, we develop a neural controlled differential equation (CDE)-based framework for identifying early PTBs in racially diverse cohorts from irregularly sampled vaginal microbial abundance data. Methods We obtained relative abundances of microbial species within vaginal microbiota using 16S rRNA sequences obtained from vaginal swabs at various stages of pregnancy. We employed a recently introduced deep learning paradigm known as ‘neural CDEs’ to predict PTBs. This method, previously unexplored, analyzes irregularly sampled microbial abundance profiles in a time-series format. Results Our framework is able to identify signatures in the temporally evolving vaginal microbiome during trimester 2 and can predict incidences of PTB (mean test set ROC–AUC = 0.81, accuracy = 0.75, F1 score = 0.71) significantly better than traditional ML classifiers, thus enabling effective early-stage PTB risk assessment. Conclusion and significance Our method is able to differentiate between term and preterm outcomes with a substantial accuracy, despite being trained using irregularly sampled microbial abundance profiles, thus overcoming the limitations of traditional time-series modeling methods.

Dynamic clustering via branched deep learning enhances personalization of stress prediction from mobile sensor data

College students experience ever-increasing levels of stress, leading to a wide range of health problems. In this context, monitoring and predicting students’ stress levels is crucial and, fortunately, made possible by the growing support for data collection via mobile devices. However, predicting stress levels from mobile phone data remains a challenging task, and off-the-shelf deep learning models are inapplicable or inefficient due to data irregularity, inter-subject variability, and the “cold start problem”. To overcome these challenges, we developed a platform named Branched CALM-Net that aims to predict students’ stress levels through dynamic clustering in a personalized manner. This is the first platform that leverages the branching technique in a multitask setting to achieve personalization and continuous adaptation. Our method achieves state-of-the-art performance in predicting student stress from mobile sensor data collected as part of the Dartmouth StudentLife study, with a ROC AUC 37% higher and a PR AUC surpassing that of the nearest baseline models. In the cold-start online learning setting, Branched CALM-Net outperforms other models, attaining an average F1 score of 87% with just 1 week of training data for a new student, which shows it is reliable and effective at predicting stress levels from mobile data.

Use of axillary ultrasound to guide breast cancer management in the genomic assay era.

Chemotherapy is conventionally offered to non-stage IV breast cancer patients with metastatic nodes. However, the RxPONDER trial showed that chemotherapy can be omitted in selected patients with 1-3 metastatic nodes if the 21-gene assay recurrence score is ≤25. We aimed to investigate if axillary ultrasound can identify this group of patients with limited nodal burden so that they can undergo upfront surgery followed by gene assay testing, to potentially avoid chemotherapy. T1-3, node positive, hormone receptor-positive and HER2-negative breast cancer patients ≥50 years old with axillary lymph node dissection (ALND) were reviewed from 2 centres. Patients with neoadjuvant chemotherapy and bilateral cancers were excluded. Number of ultrasound-detected abnormal axillary nodes, demographic and histological parameters were correlated with the number of metastatic nodes found on ALND. 138 patients were included, 59 (42.8%) and 79 (57.2%) patients had 1-3 and >3 metastatic nodes on ALND respectively. On logistic regression and ROC analysis, the number of ultrasound-detected abnormal nodes was significant (p <0.001) for predicting limited nodal burden (ROC AUC = 0.7135). Probabilities of <4 metastatic nodes with ultrasound cut-offs of 5, 6 and 8 abnormal nodes were 0.057, 0.026 and 0.005 respectively, with 100% specificity. A cut-off of ≤5 ultrasound-detected abnormal nodes can distinguish between patients with limited versus high nodal burden, with high specificity. Hence, incorporating the number of abnormal ultrasound-detected nodes into clinical practice may prove useful in guiding between upfront surgery and gene assay testing or neoadjuvant chemotherapy in this group of patients.

AI unveiled personalities: Profiling optimistic and pessimistic attitudes in Hindi dataset using transformer‐based models

AbstractBoth optimism and pessimism are intricately intertwined with an individual's inherent personality traits and people of all personality types can exhibit a wide range of attitudes and behaviours, including levels of optimism and pessimism. This paper undertakes a comprehensive analysis of optimistic and pessimistic tendencies present within Hindi textual data, employing transformer‐based models. The research represents a pioneering effort to define and establish an interaction between the personality and attitude chakras within the realm of human psychology. Introducing an innovative “Chakra” system to illustrate complex interrelationships within human psychology, this work aligns the Myers‐Briggs Type Indicator (MBTI) personality traits with optimistic and pessimistic attitudes, enriching our understanding of emotional projection in text. The study employs meticulously fine‐tuned transformer models—specifically mBERT, XLM‐RoBERTa, IndicBERT, mDeBERTa and a novel stacked mDeBERTa—trained on the novel Hindi dataset ‘मनोभाव’ (pronounced as Manobhav). Remarkably, the proposed Stacked mDeBERTa model outperforms others, recording an accuracy of 0.7785 along with elevated precision, recall, and F1 score values. Notably, its ROC AUC score of 0.7226 underlines its robustness in distinguishing between positive and negative emotional attitudes. The comparative analysis highlights the superiority of the Stacked mDeBERTa model in effectively capturing emotional attitudes in Hindi text.