Posterior Distribution Research Articles

Uncertainty quantification in multivariable regression for material property prediction with Bayesian neural networks

With the increased use of data-driven approaches and machine learning-based methods in material science, the importance of reliable uncertainty quantification (UQ) of the predicted variables for informed decision-making cannot be overstated. UQ in material property prediction poses unique challenges, including multi-scale and multi-physics nature of materials, intricate interactions between numerous factors, limited availability of large curated datasets, etc. In this work, we introduce a physics-informed Bayesian Neural Networks (BNNs) approach for UQ, which integrates knowledge from governing laws in materials to guide the models toward physically consistent predictions. To evaluate the approach, we present case studies for predicting the creep rupture life of steel alloys. Experimental validation with three datasets of creep tests demonstrates that this method produces point predictions and uncertainty estimations that are competitive or exceed the performance of conventional UQ methods such as Gaussian Process Regression. Additionally, we evaluate the suitability of employing UQ in an active learning scenario and report competitive performance. The most promising framework for creep life prediction is BNNs based on Markov Chain Monte Carlo approximation of the posterior distribution of network parameters, as it provided more reliable results in comparison to BNNs based on variational inference approximation or related NNs with probabilistic outputs.

Cyclic behavior of laminated bio-based connections with slotted-in steel plates: Genetic algorithm, deterministic neural network-based model parameter identification, and uncertainty quantification

To support more sustainable construction, this paper experimentally investigates the cyclic behavior of laminated timber (Laminated Veneer Lumber (LVL)) and glubam (Glue Laminated Bamboo) connections with slotted-in steel plates in terms of experimental test, numerical simulations and parameter identification. Experimental tests included eight different configurations: two materials (LVL and glubam), two bolt diameters (8 and 10 mm), and one or two bolts. Two different cyclic-loading protocols were applied for each type of connection: only tension and tension/compression. The observed behavior is then compared to a finite element model developed in OpenSeesPy, which takes into account factors such as sliding, contact, pinching, cyclic stiffness, and strength degradation. To identify the best set of parameters for the model, three different approaches are considered: genetic algorithm, fast deterministic neural network, and probabilistic Bayesian method. First, the model identification is carried out by means of a genetic algorithm-based optimization. The parameter-identification results are evaluated in terms of elastic stiffness, yielding point, and ductility. Next, a sensitivity analysis is performed to determine the significance of the parameters, and an innovative approach combining neural network and sensitivity analysis is proposed for fast and preliminary parameter identification. Then, probabilistic Bayesian identification is employed to calculate the posterior distribution of the model parameters identified and the confidence bounds of the estimated response. Finally, different model identification parameters are compared and suggestions for algorithm selection are provided.

First report of leaf spot on Daphniphyllum macropodum caused by Neopestalotiopsis clavispora in China.

Daphniphyllum macropodum Miq., an evergreen arbor, is widely cultivated in southern China for its ornamental and medicinal value (Su et al. 2013). In October 2019, a severe leaf spot was observed on D. macropodum in Jinggangshan National Nature Reserve in Ji'an city, Jiangxi, China (114°06'23″E, 26°32'25″N). The plants were about 15 years old, and the disease incidence was estimated to be 15% (4/26 plants). The disease primarily appeared as small black spots on the leaves. At the late stage, the spots enlarged and coalesced into regular or irregular gray necrotic lesions with dark margins. We collected five samples per plant and total 20 samples were collected to isolated the pathogen strains. The margin of the diseased tissues was cut into 5 mm × 5 mm pieces; surface disinfected with 70% ethanol and 2% NaOCl for 30 s and 60 s, respectively; and rinsed thrice with sterile water. Tissues were placed on potato dextrose agar (PDA) and incubated at 25°C in the dark. Pure cultures were obtained by single-spore isolation method, and the representative isolates, JRM3, JRM6, and JRM8 were used for morphological studies and phylogenetic analyses. The colonies of three isolates grown on PDA were white, cottony, and flocculent, contained undulate edges with dense aerial mycelium on the surface at 25 °C. Conidiomata was black conidial masses on PDA. Conidia were 5-celled, clavate to fusiform, smooth, 19.3 to 24.4 long × 6.1 to 8.6 μm wide (n = 50). The 3 median cells were dark brown to olivaceous, the central cell was darker than the other 2 cells, and the basal and apical cells were hyaline. All conidia developed one basal appendage (3.4 to 8.3 μm long; n = 50), and 2 to 3 apical appendages (18 to 32 μm long; n = 50), filiform. The morphological characteristics of the isolates are comparable with those of the genus Neopestalotiopsis (Maharachchikumbura et al. 2014). The internal transcribed spacer (ITS) regions, β-tubulin 2 (TUB2) and translation elongation factor 1-alpha (TEF1-α) were amplified from genomic DNA for the three isolates using primers ITS1/ITS4, T1/Bt-2b, EF1-728F/EF-2 (Maharachchikumbura et al. 2014), respectively. The sequences of the isolates were submitted to GenBank (ITS, OQ372202 to OQ372204; TUB2, OQ390129 to OQ390131; TEF1-α, OQ390126 to OQ390128). A maximum likelihood and Bayesian posterior probability analyses using IQtree v. 1.6.8 and Mr. Bayes v. 3.2.6 with the concatenated sequences placed JRM3, JRM6, and JRM8 in the clade of N. clavispora. Based on the multi-locus phylogeny and morphology, three isolates were identified as N. clavispora. To confirm pathogenicity, eight healthy 10-year-old D. macropodum plants growing in the field were chosen, and 4 leaves per plant were wounded with a sterile needle and inoculated with 10 μL conidial suspension per leaf (106 conidia/ml). Eight plants inoculated with sterile water were used as control. All the inoculated leaves were covered with plastic bags to maintian a humidity environment for 2 days. The leaves inoculated with conidial suspension showed similar symptoms to those observed in the field, whereas control leaves were asymptomatic for 10 days. The same fungus were re-isolated from the lesions, whereas no fungus was isolated from control leaves. N. clavispora was determined as the pathogen of a variety of plant diseases, including Kadsura coccinea (Xie et al. 2018), Dendrobium officinale (Cao et al. 2022), Macadamia integrifolia (Santos et al. 2019). However, this is the first report of N. clavispora infecting D. macropodum in China. This work provided crucial information for epidemiologic studies and appropriate control strategies for this newly emerging disease.

A bi-endpoint expectation-maximisation algorithm for re-estimating sample size for the time-to-event endpoint under the blind condition

Abstract The expectation-maximisation (EM) algorithm can be used to adjust the sample size for the time-to-event endpoint without unblinding. Nevertheless, censoring or unreliable initial estimates may render inconsistent estimates by the EM algorithm. To address these limitations, we propose a bi-endpoint EM algorithm that incorporates the time-to-event endpoint and another endpoint, which can encompass various endpoint types and is not limited to efficacy indicators, during the EM iterations. Additionally, we suggest 2 approaches for choosing initial estimates. The application conditions are as follows: (i) at least one endpoint’s initial estimate is reliable and (ii) the influence of this endpoint on the posterior distribution of the latent variable exceeds that of another endpoint.

Uncertainty Quantification of Process- Property-Structure Linkage for Fused Filament Fabrication Parts

Uncertainty Quantification of Process- Property-Structure Linkage for Fused Filament Fabrication Parts

A deep learning method for comparing Bayesian hierarchical models.

Bayesian model comparison (BMC) offers a principled approach to assessing the relative merits of competing computational models and propagating uncertainty into model selection decisions. However, BMC is often intractable for the popular class of hierarchical models due to their high-dimensional nested parameter structure. To address this intractability, we propose a deep learning method for performing BMC on any set of hierarchical models which can be instantiated as probabilistic programs. Since our method enables amortized inference, it allows efficient re-estimation of posterior model probabilities and fast performance validation prior to any real-data application. In a series of extensive validation studies, we benchmark the performance of our method against the state-of-the-art bridge sampling method and demonstrate excellent amortized inference across all BMC settings. We then showcase our method by comparing four hierarchical evidence accumulation models that have previously been deemed intractable for BMC due to partly implicit likelihoods. Additionally, we demonstrate how transfer learning can be leveraged to enhance training efficiency. We provide reproducible code for all analyses and an open-source implementation of our method. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Clinical outcomes in patients with piperacillin/ tazobactam-non-susceptible but ceftriaxone-susceptible E. coli or K. pneumoniae bloodstream infection.

A small proportion of Escherichia coli and Klebsiella pneumoniae demonstrate in vitro non-susceptibility to piperacillin/tazobactam but retain susceptibility to ceftriaxone. Uncertainty remains regarding how best to treat these isolates. We sought to compare clinical outcomes between patients with piperacillin/tazobactam-non-susceptible but ceftriaxone-susceptible E. coli or K. pneumoniae bloodstream infection receiving definitive therapy with ceftriaxone versus an alternative effective antibiotic. We retrospectively identified patients with a positive blood culture for piperacillin/tazobactam-non-susceptible but ceftriaxone-susceptible E. coli or K. pneumoniae between 1 January 2013 and 31 December 2022. Patients were divided into one of two definitive treatment groups: ceftriaxone or alternative effective antibiotic. Our primary outcome was a composite of 90 day all-cause mortality, hospital readmission, or recurrence of infection. We used Cox proportional hazards models to compare time with the composite outcome between groups. Sixty-two patients were included in our analysis. Overall, median age was 63 years (IQR 49.5-71.0), the most common source of infection was intra-abdominal (25/62; 40.3%) and the median total duration of therapy was 12.0 days (IQR 9.0-16.8). A total of 9/22 (40.9%) patients in the ceftriaxone treatment group and 18/40 (45.0%) patients in the alternative effective antibiotic group met the composite endpoint. In an adjusted time-to-event analysis, there was no difference in the composite endpoint between groups (HR 0.67, 95% CI 0.30-1.50). The adjusted Bayesian posterior probability that the HR was less than or equal to 1 (i.e. ceftriaxone is as good or better than alternative therapy) was 85%. These findings suggest that ceftriaxone can be used to effectively treat bloodstream infections with E. coli or K. pneumoniae that are non-susceptible to piperacillin/tazobactam but susceptible to ceftriaxone.

A Bayesian Spatio-Temporal Modeling Approach to the Inverse Heat Conduction Problem

Abstract A wide range of science and engineering problems involve inferring the system behavior and physical models from indirect observations. These are known as inverse problems. Unlike forward problems (predictive models), inverse problems are difficult to solve due to ill-posedness arising from one or more of non-existence, non-uniqueness (existence of many solutions), or instability (small perturbations in the measurements emerging from noise could lead to large changes in the solution). Therefore, direct inversion of the inverse problem in complex real-world systems is impractical. Several modeling approaches have been developed and adopted in the literature. However, the existing approaches do not consider spatio-temporal relationships. Meanwhile, various real systems and processes exist whose inverse problems can be comfortably modeled using spatio-temporal modeling techniques. We present a spatio-temporal inverse model considering interaction effects for modeling this type of inverse problem. We adopted a Bayesian inference framework for the subsequent parameter estimation problem and developed a Gibbs sampling algorithm to sample from the posterior distribution of the parameter of interest. Finally, we evaluate the proposed methodology through case studies on a 1D heat simulation and pool boiling experiment, using multi-sensor thermocouple measurements to reconstruct heat flux. Compared to the conventional methods, the proposed method performs better and can be adopted for solving inverse heat conduction problems. Moreover, incorporating the spatio-temporal interaction term notably improved the model's performance.

230 Using phenotypes with uncertain identification leveraging computer vision models

Abstract There has been an increase in high throughput phenotyping using sensors in livestock farming. The major obstacle when analyzing phenotypic data in studies using cameras is identifying the animals using computer vision. To overcome this problem, an analysis method that uses keypoint detection in conjunction with classification algorithms can identify animals in images. However, these algorithms generally output a prediction probability for different classes, reflecting uncertainty. In most practices, the highest probability will be assigned as the putative identification (ID). In some cases, this may introduce incorrect identification and affect downstream decisions. Here, we propose an analytical approach to incorporate all the uncertainty in the putative ID distribution into downstream inferences. We used a publicly available keypoint detection dataset of 17 keypoints detected in 3,226 images from 30 diverse thoroughbred horses. We performed feature engineering with Linear Discriminant Analysis (LDA) using 5-fold cross-validation to build horse classifiers. The best model selected had an accuracy of 88.12% to identify animals in the testing set. Afterward, we performed Monte Carlo simulations to add a phenotype that was assumed independent of the image keypoints. This generated a set of plasmodes to evaluate methods that accounted for ID uncertainty. Each simulation included different means for two groups of 15 horses. In total nine simulation scenarios were considered including varying effect sizes of group mean difference and various repeatability values (horse variance to total variance ratio). We analyzed the simulated data with three different mixed models: 1) Fixed and random effects were based on using ground truth assignment of ID, 2) Fixed and random effects were based on using the putative ID with the highest posterior probability from the best classifier, 3) Finally, prediction probabilities were used to account for putative ID uncertainty. As expected, the model that used ground truth ID had the highest power to detect group differences, and it had the lowest root mean square error (rMSE) to estimate variance components across all scenarios. The linear model that used highest posterior probability putative ID, showed the lowest power and highest rMSE. For instance, as shown in Table 1, the mean rMSE value for group mean differences in the model using ground truth is 0.01 and the model using highest posterior probabilities is 0.13. The model that incorporated ID uncertainty had intermediate properties, showing a rMSE of 0.03 for group differences. Overall, incorporating uncertainty in Animal ID results in improved performance in ultimate test statistics used in animal studies. This is very important in breeding and precision livestock farming contexts where animal identification is challenging. We will further investigate this.

62 Implications of factor analyses and Bayesian network learning to develop composite traits of size attributes in developing beef heifers

Abstract Advancements in phenotypic data collection methods provide opportunities to capture complex biological traits more efficiently. Even so, interpreting those phenotypes and using them in selection decisions is challenging in beef cattle production. Objectives of this study were to 1) develop composite phenotype(s) of economic importance in beef heifers given attributes of their body conformation and ultrasound-based measures of reproductive and carcass traits, and 2) identify associated genomic regions under selection using the structure equation modeling (SEM) approach. In this study, we hypothesized that the interplay of factor analysis and Bayesian Network Learning (BNL) approaches would better define the complex composite traits for size and carcass-related traits in beef heifers. Phenotypic data of admixed beef heifers (n = 336) for different reproductive, body conformational, and carcass-associated phenotypic traits (n = 16) were collected. Using exploratory and confirmatory factor analyses, two latent variables explaining 14 phenotypic traits were identified, herein called Body Size (BS; n = 7 traits represented) and Body Composition (BC; n = 7 traits represented). To efficiently model them in genome-wide association studies (GWAS), fixed effects of data collection year, dam age, and primary breed group as well as causal network structure identified through BNL were used within SEM model of GWAS. Body Size was directly or indirectly contributing to BC based on BNL structure. Given the data-driven approach and novelty of the data, BayesCπ model was applied with the sliding window of 1 Mb to avoid noise and linkage disequilibrium implications. Regions capturing large amounts of genetic variance (Posterior Probability of Association, PPA ≥ 0.8) were mapped to their closest gene for enrichment analysis. Twenty-four different genes were identified associated with both composite phenotypes. Functional enrichment analysis of significantly involved pathways (FDR < 0.05) indicated that genes from the HERC family (HERC3, HERC6, and HERC5) were found involved in cellular growth, and those from desmosomes (DC2, and DC3) were involved in morphogenesis like BS. Genes related to energy metabolism like IMPAD1, and PIGY, along with genes associated with traits like muscular development, and other carcass traits (i.e., FAM184B, NCAPG, and LCORL), were also found associated with BS and BC. Heritability ranged from 0.551 ± 0.048 to 0.897 ± 0.145 for BS and BC, respectively. The combined heritability of reproductive and carcass traits (BC) was much greater than the heritability of any related individual trait reported to date and was comparable for individual BS-related traits. Given this, we conclude that the implications of factor analyses with BNL can provide better biological insights for improved genomic selection in beef cattle, even for traits of low heritability.

The complete chloroplast genome of Oxytropis ochrocephala Bunge 1874 (Fabaceae) and its phylogenetic analysis

This study presents the first-ever complete chloroplast (cp) genome sequence of Oxytropis ochrocephala Bunge 1874, a member of the Fabaceae family. The cp genome spans 126,996 base pairs and includes 109 genes, comprising 76 protein-coding genes, 29 tRNA genes, and four rRNA genes. Notably, the genome lacks an inverted repeat (IR) region. Additionally, we constructed phylogenetic trees for 34 species within Trib. Galegeae, employing both maximum likelihood (ML) and Bayesian inference (BI) methods. These analyses robustly support the monophyly of the Oxytropis species, evidenced by high bootstrap values (BP = 100) and posterior probabilities (PP = 1). Within this clade, O. ochrocephala exhibits a sister relationship with other Oxytropis species. Our findings provide valuable insights into the genetic makeup and evolutionary relationships of O. ochrocephala within the Galegeae tribe.

Bayesian inference of vorticity in unbounded flow from limited pressure measurements

We study the instantaneous inference of an unbounded planar flow from sparse noisy pressure measurements. The true flow field comprises one or more regularized point vortices of various strength and size. We interpret the true flow's measurements with a vortex estimator, also consisting of regularized vortices, and attempt to infer the positions and strengths of this estimator assuming little prior knowledge. The problem often has several possible solutions, many due to a variety of symmetries. To deal with this ill posedness and to quantify the uncertainty, we develop the vortex estimator in a Bayesian setting. We use Markov-chain Monte Carlo and a Gaussian mixture model to sample and categorize the probable vortex states in the posterior distribution, tailoring the prior to avoid spurious solutions. Through experiments with one or more true vortices, we reveal many aspects of the vortex inference problem. With fewer sensors than states, the estimator infers a manifold of equally possible states. Using one more sensor than states ensures that no cases of rank deficiency arise. Uncertainty grows rapidly with distance when a vortex lies outside of the vicinity of the sensors. Vortex size cannot be reliably inferred, but the position and strength of a larger vortex can be estimated with a much smaller one. In estimates of multiple vortices their individual signs are discernible because of the nonlinear coupling in the pressure. When the true vortex state is inferred from an estimator of fewer vortices, the estimate approximately aggregates the true vortices where possible.

A Practical probabilistic history matching framework for characterizing fracture network parameters of shale gas reservoirs

Simulating fluid flow in complex fracture systems encountered in subsurface systems, such as those encountered in tight or shale gas reservoirs, presents considerable challenges. Sophisticated numerical techniques and characterization approaches are required to integrate static and dynamic data from multiple sources to account for the intricate interplay between fractures and the surrounding rock matrix. A novel workflow using an indicator-based probability perturbation method (PPM) is applied to characterize uncertain discrete fracture network parameters. The posterior probability distributions of primary and secondary fracture transmissivity, aperture, length, height, and intensity of the secondary fracture are estimated according to the production (flow) histories. A case study based on the Horn River shale gas reservoir is presented. A pilot point scheme is formulated to update the distributions of P32L. The forward modelling entails upscaling each fracture model into a dual-porosity dual-permeability model and performing multiphase flow simulation. This approach produces an ensemble of history-matched discrete fracture and upscaled models by incorporating static and dynamic data. The results demonstrate the utility of the developed method for estimating secondary fracture parameters, which are not inferrable from other static information alone. The inference of both primary and secondary fractures offers a more comprehensive characterization of the fracture networks in tight or shale gas reservoirs. Integrating a pilot point parameterization technique and a sequential simulation approach into the PPM framework presents significant novelties in contributing to a comprehensive and effective method of simulating fluid flow in complex fracture systems, addressing the challenges associated with non-linear relationships between fracture model parameters and the corresponding flow response.

Online detection of the incidence via transfer learning

AbstractThe counting process has abundant applications in reality, and Poisson process monitoring actually has received extensive attention and research. However, conventional methods experience poor performance when shifts appears early and only small number of historical observations in Phase I can be used for estimation. To overcome it, we creatively propose a new online monitoring algorithm under the transfer learning framework, which utilizes the information from observations of additional data sources so that the target process can be described better. By making the utmost of the somewhat correlated data from other domains, which is measured by a bivariate Gamma distributed statistic presented by us, the explicit properties (e.g., posterior probability mass function, posterior expectation, and posterior variance) are also strictly proved. Furthermore, based on the above theoretical results, we design two computationally efficient control schemes in Phase II, that is a control chart based on the cumulative distribution function for large shifts and an exponentially weighted moving average control chart for small shifts. For a better understanding of the more practical applications and transferability matter, we provide some optimal values for parameter setting. Extensive numerical simulations and a case of skin cancer incidence in America verify the superiorities of our approach.

Equivalence between constrained optimal smoothing and Bayesian estimation

In this paper, we extend the correspondence between Bayesian estimation and optimal smoothing in a Reproducing Kernel Hilbert Space (RKHS) by adding convex constraints to the problem. Through a sequence of approximating Hilbertian subspaces and a discretised model, we prove that the Maximum a posteriori (MAP) of the posterior distribution is exactly the optimal constrained smoothing function in the RKHS. This paper can be read as a generalisation of the paper Kimeldorf and Wahba (1970), where it is proved that the optimal smoothing solution is the mean of the posterior distribution. Synthetic and real data studies confirm the correspondence established in this paper.

Smoking Cessation After Initial Treatment Failure With Varenicline or Nicotine Replacement

Most people who smoke do not quit on their initial attempt. To determine the best subsequent strategy for nonabstinence following initial treatment with varenicline or combined nicotine replacement therapy (CNRT). Using a double-blind, placebo-controlled, sequential multiple assignment randomized trial, 490 volunteers were randomized to receive 6 weeks of varenicline or CNRT. After 6 weeks, nonabstainers were rerandomized to continue, switch, or increase medication dosage for 6 additional weeks. The study was conducted from June 2015 through October 2019 in a Texas tobacco treatment clinic. The initial treatment was 2 mg/d of varenicline or the combined replacement therapy of a 21-mg patch plus 2-mg lozenge. The rerandomized participants either continued with their initial therapies, switched between varenicline and CNRT, or increased dosages either to 3-mg or more of varenicline or to a 42-mg patch and lozenges. All received weekly brief counseling. Biochemically verified 7-day point prevalence abstinence at the end of treatment at 12 weeks. The 490 randomized participants (210 female [43%], 287 non-Hispanic White [58%], mean age, 48.1 years) smoked an average of 20 cigarettes per day. After the first phase, 54 participants in the CNRT group were abstinent and continued their therapy; of the 191 who were not abstinent, 151 were rerandomized, and the 40 who did not return for rerandomization were assigned to continue their initial CNRT condition in phase 2. The end-of-treatment abstinence rate for the 191 phase 1 nonabstainers was 8% (95% credible interval [CrI], 6% to 10%) for the 90 (47%) who continued at the dosage condition, 14% (CrI, 10% to 18%) for the 50 (33%) who increased their dosage, and 14% (95% CrI, 10% to 18%) for the 51 (34%) who switched to varenicline (absolute risk difference [RD], 6%; 95% CrI, 6% to 11%) with more than 99% posterior probability that either strategy conferred benefit over continuing the initial dosage. After the first phase, 88 participants in the varenicline group were abstinent and continued their therapy; of the 157 who were not abstinent, 122 were rerandomized and 35 who did not return for rerandomization were assigned to continue with the varenicline condition. The end-of-treatment abstinence rate for the 157 phase 1 nonabstainers was 20% (95% CrI, 16% to 26%) for the 39 (32%) who increased their varenicline dosage, 0 (95% CrI, 0 to 0) for the 41 (34%) who switched CNRT, and 3% (95% CrI, 1% to 4%) for the 77 (49%) who were assigned to the continued varenicline condition (absolute RD, -3%; 95% CrI, -4% to -1%) with more than 99% posterior probability that continuing varenicline at the initial dosage was worse than switching to a higher dosage. Furthermore, increasing the varenicline dosage had an absolute RD of 18% (95% CrI, 13% to 24%) and a more than 99% posterior probability of conferring benefit. The secondary outcome of continuous abstinence at 6 months indicated that only increased dosages of the CNRT and varenicline provided benefit over continuation of the initial treatment dosages. For individuals who smoked but did not achieve abstinence after treatment with varenicline, increasing the dosage enhanced abstinence vs continuing, whereas for nonabstainers initially treated with CNRT, a dosage increase or switch to varenicline enhanced abstinence and may be viable rescue strategies. ClinicalTrials.gov Identifier: NCT02271919.

Abstract PO5-09-03: Multi-tissue transcriptome-wide association studies identified genes for intrinsic subtypes of breast cancer

Abstract Background: Though several genome-wide association studies (GWAS) of breast cancer (BC) have identified common variants which differ between intrinsic subtypes, the genes through which these variants act through to impact BC risk have not been fully established. Furthermore, transcriptome-wide association studies (TWAS) have thus far been primarily employed to identify genes associated with overall BC risk, while overlooking how the influence of common variation on gene expression may contribute to subtype-specific differences. Methods: In this study, we performed two complementary multi-tissue TWASs for each of the following BC intrinsic subtypes: Luminal A-like, Luminal B-like, Luminal B/HER2-negative-like, HER2-enriched-like, Triple Negative BC. These two approaches included 1) an expression-based approach that collated TWAS signals from expression quantitative trait loci (eQTLs) across multiple tissues using the Aggregated Cauchy Association Test (ACAT) and 2) a splicing-based approach that utilized two applications of ACAT to collate splicing QTLs (sQTLs) for a given gene in a tissue and then across tissues. To perform these two TWASs, we utilized e/sQTL models trained in 11 tissues from the Genotype-Tissue Expression Project including breast, ovary, uterus, vagina, EBV-transformed lymphocytes, whole blood, spleen, liver, subcutaneous adipose, visceral adipose, and cell-cultured fibroblasts. GWAS summary statistics were previously generated from 133,384 BC cases and 113,789 controls who were participants in the Breast Cancer Association Consortium (BCAC). We further performed our TWAS while conditioning e/sQTL effect sizes on nearby GWAS index SNPs. Additionally, we utilized gene-based fine-mapping of eQTLs and sQTL to identify candidate causal genes for each intrinsic subtype. Results: Overall, we identified 164 genes in 69 loci that were associated with Luminal A-like, 19 genes in 9 loci with Luminal B-like, 18 genes in 11 loci with Luminal B/HER2-negative-like, 10 genes in 7 loci with HER2-enriched-like, and 29 genes in 12 loci with TNBC. Among these genes, 17 genes had not been reported in previous TWAS of BC, and 140 genes, 1 gene, 2 genes, 2 genes, and 16 genes were uniquely associated with each of the intrinsic subtypes, respectively. Additionally, we identified one gene associated with Luminal A-like and one gene with Luminal B-like BC that were each in a locus located at least 1.4 Mb from published GWAS hits. Furthermore, we identified 106, 11, 10, 5, and 21 candidate causal genes for each of the intrinsic subtypes, respectively, that had a posterior inclusion probability of 0.9 in at least one e/sQTL. Conclusion: In summary, our multi-tissue TWAS corroborated previous GWAS loci for overall BC risk and intrinsic subtypes, while underscoring how common variation impacts BC etiology by modulating the expression and splicing of genes in multiple tissue types. Citation Format: James Li, Julian McClellan, Haoyu Zhang, Guimin Gao, Dezheng Huo. Multi-tissue transcriptome-wide association studies identified genes for intrinsic subtypes of breast cancer [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO5-09-03.

Metal mixture exposures and serum lipid levels in childhood: the Rhea mother-child cohort in Greece.

Growing evidence suggests that cardiovascular disease develops over the lifetime, often beginning in childhood. Metal exposures have been associated with cardiovascular disease and important risk factors, including dyslipidemia, but prior studies have largely focused on adult populations and single metal exposures. To investigate the individual and joint impacts of multiple metal exposures on lipid levels during childhood. This cross-sectional study included 291 4-year-old children from the Rhea Cohort Study in Heraklion, Greece. Seven metals (manganese, cobalt, selenium, molybdenum, cadmium, mercury, and lead) were measured in whole blood using inductively coupled plasma mass spectrometry. Serum lipid levels included total cholesterol, triglycerides, high-density lipoprotein (HDL) cholesterol, and low-density lipoprotein (LDL) cholesterol. To determine the joint and individual impacts of child metal exposures (log2-transformed) on lipid levels, Bayesian kernel machine regression (BKMR) was employed as the primary multi-pollutant approach. Potential effect modification by child sex and childhood environmental tobacco smoke exposure was also evaluated. BKMR identified a positive association between the metal mixture and both total and LDL cholesterol. Of the seven metals examined, selenium (median 90.6 [IQR = 83.6, 96.5] µg/L) was assigned the highest posterior inclusion probability for both total and LDL cholesterol. A difference in LDL cholesterol of 8.22 mg/dL (95% CI = 1.85, 14.59) was observed when blood selenium was set to its 75th versus 25th percentile, holding all other metals at their median values. In stratified analyses, the positive association between selenium and LDL cholesterol was only observed among boys or among children exposed to environmental tobacco smoke during childhood. Growing evidence indicates that cardiovascular events in adulthood are the consequence of the lifelong atherosclerotic process that begins in childhood. Therefore, public health interventions targeting childhood cardiovascular risk factors may have a particularly profound impact on reducing the burden of cardiovascular disease. Although growing evidence supports that both essential and nonessential metals contribute to cardiovascular disease and risk factors, such as dyslipidemia, prior studies have mainly focused on single metal exposures in adult populations. To address this research gap, the current study investigated the joint impacts of multiple metal exposures on lipid concentrations in early childhood.

Abstract PS16-01: Comparison of an Atezolizumab monotherapy window followed by Atezolizumab and chemotherapy vs. Atezolizumab and chemotherapy alone in triple negative breast cancer (TNBC) – final analysis of the neoadjuvant neoMono trial

Abstract Background: Improvement of systemic therapy of TNBC still is a medical need. Exploratory data from the neoadjuvant GeparNuevo trial suggested a benefit from an immune checkpoint inhibitor (ICI) monotherapy window in TNBC. The neoMono trial prospectively analyzed whether the addition of a preceding Atezolizumab monotherapy window prior to Atezolizumab and chemotherapy (CTX) improves pCR rates among patients (pts) with early TNBC. In an interim analysis after 100 pCR events, patients with unselected TNBC did not show a significant benefit from an Atezolizumab monotherapy window, while an exploratory analysis suggested a highly clinically relevant benefit among pts with PD-L1 positive TNBC. Here we present the final primary endpoint analysis. Methods: NeoMono is a phase 2 randomized multicenter trial that was planned to recruit a maximum of 458 female and male pts with primary TNBC (defined as ER/PR < 10% and HER2 negative) with tumor stages cT1c – cT4d (cN0 and cN+). As the protocol mandated termination of trial recruitment based on the results of the interim analysis, the final ITT population was limited to 359 pts. PD-L1 status had to be identifiable by central pathology by means of the VENTANA PD-L1 (SP142) assay and was defined by PD-L1 expression on immune cells (IC). Neoadjuvant treatment in both study arms consisted of Atezolizumab 1200 mg every 3 weeks in addition to neoadjuvant CTX (12 x Carboplatin/Paclitaxel q1w followed by 4x Epirubicin/Cyclophosphamide q3w). Combination therapy in arm A was preceded by an Atezolizumab monotherapy window of 840 mg once two weeks prior to initiation of combination therapy, while patients in arm B received no immune monotherapy window. Study goals are to compare the efficacy of neoadjuvant CTX + Atezolizumab with versus without a two-week atezolizumab monotherapy window preceding CTX + ICI (primary endpoint: pCR) and to identify biomarkers for response and resistance through analysis of sequential tissue and liquid biopsies. The neoMono statistical design uses Bayesian posterior probabilities (uniform prior distribution) and logistic regression to analyze the primary endpoint. Results: 180 pts in arm A and 179 in arm B from 34 study sites were included in the final primary endpoint analysis. Demographics and baseline characteristics as well as drug exposure were well-balanced in both arms. Posterior mean pCR rates in the ITT population in study arms A and B were 65.7% (95% high posterior density interval (HPDI): 58.5%, 72.5%) and 69% (62.2%, 75.9%), respectively. In an exploratory analysis stratified by PD-L1 IC status (negative: < 1% versus positive: ≥ 1%), pCR rates in arm A were 91.5% in the PD-L1 IC-positive group and 56.1% in the PD-L1 IC-negative group, the corresponding pCR rates in arm B were 82.2% and 64.5%, respectively. In a multivariate analysis of the ITT population including tumor size, nodal status, tumor grade, age and PD-L1 status, the odds ratio for achieving a pCR was 4.77 (p < 0.001) for PD-L1-positive and 2.36 (p=0.023) for grade 3 tumors. In an exploratory analysis including HER2 status, odds for achieving a pCR were significantly higher for patients with HER2-negative (IHC 0) vs. HER2-low tumors (OR 1.73, p=0.036). No new safety signals were observed. Conclusion: The final primary endpoint analysis of the neoMono trial demonstrated the highest pCR rates ever reported in a phase II/III trial in TNBC. While a significant impact of an ICI monotherapy window on the pCR rate after combination of CTX + ICI in an unselected ITT population could not be demonstrated, neoMono indicates for the first time in a randomized prospective setting that patients with immune active TNBC might derive particular benefit from a preceding ICI monotherapy window. The results of the neoMono trial are mainly justifying the conduction of a confirmative trial in immune active TNBC. However, our results underscore the potential role of pre-therapeutic ICI monotherapy window as part of future therapeutic concepts in TNBC. Citation Format: Hans-Christian Kolberg, Johannes Schumacher, Ramona Erber, Michael Braun, Peter A. Fasching, Eva-Maria Grischke, Christian Schem, Michael P. Lux, Mustafa Deryal, Oliver Hoffmann, Bernhard Heinrich, Georg Kunz, Kristina Lübbe, Petra Krabisch, Arndt Hartmann, Philip Raeth, Sabine Kasimir-Bauer, Cornelia Kolberg-Liedtke. Comparison of an Atezolizumab monotherapy window followed by Atezolizumab and chemotherapy vs. Atezolizumab and chemotherapy alone in triple negative breast cancer (TNBC) – final analysis of the neoadjuvant neoMono trial [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PS16-01.

Unveiling potential therapeutic targets for diabetes-induced frozen shoulder through Mendelian randomization analysis of the human plasma proteome

IntroductionThis study aimed to assess the causal relationship between diabetes and frozen shoulder by investigating the target proteins associated with diabetes and frozen shoulder in the human plasma proteome through...