FRAMM: Fair ranking with missing modalities for clinical trial site selection.

The underrepresentation of gender, racial, and ethnic minorities in clinical trials is a problem undermining the efficacy of treatments on minorities and preventing precise estimates of the effects within these subgroups. We propose FRAMM, a deep reinforcement learning framework for fair trial site selection to help address this problem. We focus on two real-world challenges: the data modalities used to guide selection are often incomplete for many potential trial sites, and the site selection needs to simultaneously optimize for both enrollment and diversity. To address the missing data challenge, FRAMM has a modality encoder with a masked cross-attention mechanism for bypassing missing data. To make efficient trade-offs, FRAMM uses deep reinforcement learning with a reward function designed to simultaneously optimize for both enrollment and fairness. We evaluate FRAMM using real-world historical clinical trials and show that it outperforms the leading baseline in enrollment-only settings while also greatly improving diversity.

Optimizing Iodine Enrichment through Induced-Fit Transformations in a Flexible Ag(I)-Organic Framework: From Accelerated Adsorption Kinetics to Record-High Storage Density.

Efficient capture and storage of radioactive I2 is a prerequisite for developing nuclear power but remains a challenge. Here, two flexible Ag-MOFs (FJI-H39 and 40) with similar active sites but different pore sizes and flexibility are prepared; both of them can capture I2 with excellent removal efficiencies and high adsorption capacities. Due to the more flexible pores, FJI-H39 not only possesses the record-high I2 storage density among all the reported MOFs but also displays a very fast adsorption kinetic (124 times faster than FJI-H40), while their desorption kinetics are comparable. Mechanistic studies show that FJI-H39 can undergo induced-fit transformations continuously (first contraction then expansion), making the adsorbed iodine species enrich near the Ag(I) nodes quickly and orderly, from discrete I- anion to the dense packing of various iodine species, achieving the very fast adsorption kinetic and the record-high storage density simultaneously. However, no significant structural transformations caused by the adsorbed iodine are observed in FJI-H40. In addition, FJI-H39 has excellent stability/recyclability/obtainability, making it a practical adsorbent for radioactive I2. This work provides a useful method for synthesizing practical radioactive I2 adsorbents.

Water-stable metal-organic frameworks (MOFs): rational construction and carbon dioxide capture.

Metal-organic frameworks (MOFs) are considered to be a promising porous material due to their excellent porosity and chemical tailorability. However, due to the relatively weak strength of coordination bonds, the stability (e.g., water stability) of MOFs is usually poor, which severely inhibits their practical applications. To prepare water-stable MOFs, several important strategies such as increasing the bonding strength of building units and introducing hydrophobic units have been proposed, and many MOFs with excellent water stability have been prepared. Carbon dioxide not only causes a range of climate and health problems but also is a by-product of some important chemicals (e.g., natural gas). Due to their excellent adsorption performances, MOFs are considered as a promising adsorbent that can capture carbon dioxide efficiently and energetically, and many water-stable MOFs have been used to capture carbon dioxide in various scenarios, including flue gas decarbonization, direct air capture, and purified crude natural gas. In this review, we first introduce the design and synthesis of water-stable MOFs and then describe their applications in carbon dioxide capture, and finally provide some personal comments on the challenges facing these areas.

Identification of a novel m6A-related lncRNAs signature and immunotherapeutic drug sensitivity in pancreatic adenocarcinoma.

BACKGROUND: Pancreatic adenocarcinoma (PDAC) ranks as the fourth leading cause for cancer-related deaths worldwide. N6-methyladenosine (m6A) and long non-coding RNAs (lncRNAs) are closely related with poor prognosis and immunotherapeutic effect in PDAC. The aim of this study is to construct and validate a m6A-related lncRNAs signature and assess immunotherapeutic drug sensitivity in PDAC. METHODS: RNA-seq data for 178 cases of PDAC patients and 167 cases of normal pancreatic tissue were obtained from TCGA and GTEx databases, respectively. A set of 21 m6A-related genes were downloaded based on the previous report. Co-expression network was conducted to identify m6A-related lncRNAs in PDAC. Cox analyses and least absolute shrinkage and selection operator (Lasso) regression model were used to construct a risk prognosis model. The relationship between signature genes and immune function was explored by single-sample GSEA (ssGSEA). The tumor immune dysfunction and exclusion (TIDE) score and tumor mutation burden (TMB) were utilized to evaluate the response to immunotherapy. Furthermore, the expression levels of 4 m6A-related lncRNAs on PDAC cell lines were measured by the quantitative real-time PCR (qPCR). The drug sensitivity between the high- and low-risk groups was validated using PDAC cell lines by Cell-Counting Kit 8 (CCK8). RESULTS: The risk prognosis model was successfully constructed based on 4 m6A-related lncRNAs, and PDAC patients were divided into the high- and low-risk groups. The overall survival (OS) of the high-risk groups was more unfavorable compared with the low-risk groups. Receiver operating characteristic (ROC) curves demonstrated that the risk prognosis model reasonably predicted the 2-, 3- and 5-year OS of PDAC patients. qPCR analysis confirmed the decreased expression levels of 4 m6A-related lncRNAs in PDAC cells compared to the normal pancreatic cells. Furthermore, CCK8 assay revealed that Phenformin exhibited higher sensitivity in the high-risk groups, while Pyrimethamine exhibited higher sensitivity in the low-risk groups. CONCLUSION: The prognosis of patients with PDAC were well predicted in the risk prognosis model based on m6A-related lncRNAs, and selected immunotherapy drugs have potential values for the treatment of pancreatic cancer.

Self-Supervised Electroencephalogram Representation Learning for Automatic Sleep Staging: Model Development and Evaluation Study.

Background: Deep learning models have shown great success in automating tasks in sleep medicine by learning from carefully annotated electroencephalogram (EEG) data. However, effectively using a large amount of raw EEG data remains a challenge. Objective: In this study, we aim to learn robust vector representations from massive unlabeled EEG signals, such that the learned vectorized features (1) are expressive enough to replace the raw signals in the sleep staging task, and (2) provide better predictive performance than supervised models in scenarios involving fewer labels and noisy samples. Methods: We propose a self-supervised model, Contrast with the World Representation (ContraWR), for EEG signal representation learning. Unlike previous models that use a set of negative samples, our model uses global statistics (ie, the average representation) from the data set to distinguish signals associated with different sleep stages. The ContraWR model is evaluated on 3 real-world EEG data sets that include both settings: at-home and in-laboratory EEG recording. Results: ContraWR outperforms 4 recently reported self-supervised learning methods on the sleep staging task across 3 large EEG data sets. ContraWR also supersedes supervised learning when fewer training labels are available (eg, 4% accuracy improvement when less than 2% of data are labeled on the Sleep EDF data set). Moreover, the model provides informative, representative feature structures in 2D projection. Conclusions: We show that ContraWR is robust to noise and can provide high-quality EEG representations for downstream prediction tasks. The proposed model can be generalized to other unsupervised physiological signal learning tasks. Future directions include exploring task-specific data augmentations and combining self-supervised methods with supervised methods, building upon the initial success of self-supervised learning reported in this study.

Prediction of prognosis, immune infiltration, and personalized treatment of hepatocellular carcinoma by analysis of cuproptosis-related long noncoding RNAs and verification in vitro.

Background: The correlations between cuproptosis and long noncoding RNAs (lncRNAs) with the tumor microenvironment (TME), immunotherapy, and some other characteristics of hepatocellular carcinoma (HCC) remain unclear. Methods: Sixteen cuproptosis regulators and 356 cuproptosis-related lncRNAs (CRLnc) were identified from 374 HCC profiles in The Cancer Genome Atlas (TCGA) database. Six differentially expressed CRLnc were selected, and a prognostic risk model based on the CRLnc signature (CRLncSig) was constructed. The prognostic power of the model was verified. Moreover, a cuproptosis-related gene cluster (CRGC) was generated based on six lncRNAs and differentially expressed genes. The relationship between immune cell infiltration in the TME, immunotherapy, CRLncSig, and CRGC was demonstrated through various algorithms, Tumor Immune Dysfunction and Exclusion (TIDE), tumor mutational burden (TMB), etc. Potential drugs and sensitivity to those agents were evaluated for the risk model. LncRNA AL158166.1 was selected and verified in HCC tissues and cell lines, the impact of its knockdown and overexpression in HCC cells was examined, and the copper (Cu) concentration and the cuproptosis-related gene expression were detected. Results: A CRLncSig prognostic risk model with good predictive ability was constructed. The low-risk group had a longer overall survival (OS), lower tumor purity, more extensive immune cell infiltration, higher immune score, enrichment in immune-activated pathways, and more positive response to immunotherapy versus the high-risk group. CRGC-B exhibited the best OS and the lowest tumor stage; the immune cell infiltration analysis was similar to the low-risk group in CRLncSig. CRGC-B belonged to the "immune-high" group of the TME. The low-risk group had a higher TIDE score and susceptibility to antitumor drugs. The lncRNA AL158166.1 had the highest hazard ratio. The levels of AL158166.1 were higher in HCC tissues versus healthy tissues. Knockdown of AL158166.1 could lead to an increase in intracellular Cu concentration, induce DLAT low expression, and inhibit the proliferation and migration of HCC cells, whereas overexpression of AL158166.1 exerted the reverse effect. Conclusion: Overall, a new CRLncSig prognostic risk model and a cuproptosis-related molecular signature were constructed and evaluated. The model and signature were associated with the prognosis, immune infiltration, and immunotherapy of HCC. Inhibiting the lncRNA AL158166.1 may induce cuproptosis and showed potential for the inhibition of tumors. Evaluation of the CRLnc, CRLncSig, and CRGC may enhance our understanding of the TME, determine the effectiveness of immunotherapy, and act as a marker for the prognosis of HCC.

Synthesize high-dimensional longitudinal electronic health records via hierarchical autoregressive language model.

Synthetic electronic health records (EHRs) that are both realistic and privacy-preserving offer alternatives to real EHRs for machine learning (ML) and statistical analysis. However, generating high-fidelity EHR data in its original, high-dimensional form poses challenges for existing methods. We propose Hierarchical Autoregressive Language mOdel (HALO) for generating longitudinal, high-dimensional EHR, which preserve the statistical properties of real EHRs and can train accurate ML models without privacy concerns. HALO generates a probability density function over medical codes, clinical visits, and patient records, allowing for generating realistic EHR data without requiring variable selection or aggregation. Extensive experiments demonstrated that HALO can generate high-fidelity data with high-dimensional disease code probabilities closely mirroring (above 0.9 R2 correlation) real EHR data. HALO also enhances the accuracy of predictive modeling and enables downstream ML models to attain similar accuracy as models trained on genuine data.

Salmonella enterica serovar Paratyphi A-induced immune response in Caenorhabditis elegans depends on MAPK pathways and DAF-16.

Salmonella enterica serovar Paratyphi A (S. Paratyphi A) is a pathogen that can cause enteric fever. According to the recent epidemic trends of typhoid fever, S. Paratyphi A has been the major important causative factor in paratyphoid fever. An effective vaccine for S. Paratyphi A has not been developed, which made it a tricky public health concern. Until now, how S. Paratyphi A interacts with organisms remain unknown. Here using lifespan assay, we found that S. Paratyphi A could infect Caenorhabditis elegans (C. elegans) at 25°C, and attenuate thermotolerance. The immune response of C. elegans was mediated by tir-1, nsy-1, sek-1, pmk-1, mpk-1, skn-1, daf-2 and daf-16, suggesting that S. Paratyphi A could regulate the MAPK and insulin pathways. Furthermore, we observed several phenotypical changes when C. elegans were fed S. Paratyphi A, including an accelerated decline in body movement, reduced the reproductive capacity, shortened spawning cycle, strong preference for OP50, arrested pharyngeal pumping and colonization of the intestinal lumen. The virulence of S. Paratyphi A requires living bacteria and is not mediated by secreting toxin. Using hydrogen peroxide analysis and quantitative RT-PCR, we discovered that S. Paratyphi A could increase oxidative stress and regulate the immune response in C. elegans. Our results sheds light on the infection mechanisms of S. Paratyphi A and lays a foundation for drugs and vaccine development.

Nucleolar HEAT Repeat Containing 1 Up-regulated by the Mechanistic Target of Rapamycin Complex 1 Signaling Promotes Hepatocellular Carcinoma Growth by Dominating Ribosome Biogenesis and Proteome Homeostasis.

BACKGROUND & AIMS: Hyperactivation of ribosome biogenesis leads to hepatocyte transformation and plays pivotal roles in hepatocellular carcinoma (HCC) development. We aimed to identify critical ribosome biogenesis proteins that are overexpressed and crucial in HCC progression. METHODS: HEAT repeat containing 1 (HEATR1) expression and clinical correlations were analyzed using The Cancer Genome Atlas and Gene Expression Omnibus databases and further evaluated by immunohistochemical analysis of an HCC tissue microarray. Gene expression was knocked down by small interfering RNA. HEATR1-knockdown cells were subjected to viability, cell cycle, and apoptosis assays and used to establish subcutaneous and orthotopic tumor models. Chromatin immunoprecipitation and quantitative polymerase chain reaction were performed to detect the association of candidate proteins with specific DNA sequences. Endogenous coimmunoprecipitation combined with mass spectrometry was used to identify protein interactions. We performed immunoblot and immunofluorescence assays to detect and localize proteins in cells. The nucleolus ultrastructure was detected by transmission electron microscopy. Click-iT (Thermo Fisher Scientific) RNA imaging and puromycin incorporation assays were used to measure nascent ribosomal RNA and protein synthesis, respectively. Proteasome activity, 20S proteasome foci formation, and protein stability were evaluated in HEATR1-knockdown HCC cells. RESULTS: HEATR1 was the most up-regulated gene in a set of ribosome biogenesis mediators in HCC samples. High expression of HEATR1 was associated with poor survival and malignant clinicopathologic features in patients with HCC and contributed to HCC growth in vitro and in vivo. HEATR1 expression was regulated by the transcription factor specificity protein 1, which can be activated by insulin-like growth factor 1-mammalian target of rapamycin complex 1 signaling in HCC cells. HEATR1 localized predominantly in the nucleolus, bound to ribosomal DNA, and was associated with RNA polymerase I transcription/processing factors. Knockdown of HEATR1 disrupted ribosomal RNA biogenesis and impaired nascent protein synthesis, leading to reduced cytoplasmic proteasome activity and inhibitory-κB/nuclear factor-κB signaling. Moreover, HEATR1 knockdown induced nucleolar stress with increased nuclear proteasome activity and inactivation of the nucleophosmin 1-MYC axis. CONCLUSIONS: Our study revealed that HEATR1 is up-regulated by insulin-like growth factor 1-mammalian target of rapamycin complex 1-specificity protein 1 signaling in HCC and functions as a crucial regulator of ribosome biogenesis and proteome homeostasis to promote HCC development.

Synthesize Extremely High-dimensional Longitudinal Electronic Health Records via Hierarchical Autoregressive Language Model.

Synthetic electronic health records (EHRs) that are both realistic and preserve privacy can serve as an alternative to real EHRs for machine learning (ML) modeling and statistical analysis. However, generating high-fidelity and granular electronic health record (EHR) data in its original, highly-dimensional form poses challenges for existing methods due to the complexities inherent in high-dimensional data. In this paper, we propose Hierarchical Autoregressive Language mOdel (HALO) for generating longitudinal high-dimensional EHR, which preserve the statistical properties of real EHR and can be used to train accurate ML models without privacy concerns. Our HALO method, designed as a hierarchical autoregressive model, generates a probability density function of medical codes, clinical visits, and patient records, allowing for the generation of realistic EHR data in its original, unaggregated form without the need for variable selection or aggregation. Additionally, our model also produces high-quality continuous variables in a longitudinal and probabilistic manner. We conducted extensive experiments and demonstrate that HALO can generate high-fidelity EHR data with high-dimensional disease code probabilities ( d ≈ 10,000), disease code co-occurrence probabilities within a visit ( d ≈ 1,000,000), and conditional probabilities across consecutive visits ( d ≈ 5,000,000) and achieve above 0.9 R 2 correlation in comparison to real EHR data. In comparison to the leading baseline, HALO improves predictive modeling by over 17% in its predictive accuracy and perplexity on a hold-off test set of real EHR data. This performance then enables downstream ML models trained on its synthetic data to achieve comparable accuracy to models trained on real data (0.938 area under the ROC curve with HALO data vs. 0.943 with real data). Finally, using a combination of real and synthetic data enhances the accuracy of ML models beyond that achieved by using only real EHR data.

ClinicalRisk: A New Therapy-related Clinical Trial Dataset for Predicting Trial Status and Failure Reasons.

Clinical trials aim to study new tests and evaluate their effects on human health outcomes, which has a huge market size. However, carrying out clinical trials is expensive and time-consuming and often ends in no results. It will revolutionize clinical practice if we can develop an effective model to automatically estimate the status of a clinical trial and find out possible failure reasons. However, it is challenging to develop such a model because of the lack of a benchmark dataset. To address these challenges, in this paper, we first build a new dataset by extracting the publicly available clinical trial reports from ClinicalTrials.gov. The associated status of each report is treated as the status label. To analyze the failure reasons, domain experts help us manually annotate each failed report based on the description associated with it. More importantly, we examine several state-of-the-art text classification baselines on this task and find out that the unique format of the clinical trial protocols plays an essential role in affecting prediction accuracy, demonstrating the need for specially designed clinical trial classification models.

HINT: Hierarchical interaction network for clinical-trial-outcome predictions.

Clinical trials are crucial for drug development but often face uncertain outcomes due to safety, efficacy, or patient-recruitment problems. We propose the Hierarchical Interaction Network (HINT) to predict clinical trial outcomes. First, HINT encodes multi-modal data (drug molecule, target disease, trial eligibility criteria) into embeddings. Then, HINT trains knowledge-embedding modules using drug pharmacokinetic and historical trial data. Finally, a hierarchical interaction graph connects all of the embeddings to capture their interactions and predict trial outcomes. HINT was trained and validated on 1,160 phase I trials, 4,449 phase II trials, and 3,436 phase III trials. It obtained 0.665, 0.620, and 0.847 F1 scores on separate test sets of 627 phase I, 1,653 phase II, and 1,140 phase III trials, respectively. HINT significantly outperforms the best baseline method on most metrics. The benchmark dataset and codes are released at https://github.com/futianfan/clinical-trial-outcome-prediction.

SCRIB: Set-Classifier with Class-Specific Risk Bounds for Blackbox Models.

Despite deep learning (DL) success in classification problems, DL classifiers do not provide a sound mechanism to decide when to refrain from predicting. Recent works tried to control the overall prediction risk with classification with rejection options. However, existing works overlook the different significance of different classes. We introduce Set-classifier with Class-specific RIsk Bounds (SCRIB) to tackle this problem, assigning multiple labels to each example. Given the output of a black-box model on the validation set, SCRIB constructs a set-classifier that controls the class-specific prediction risks. The key idea is to reject when the set classifier returns more than one label. We validated SCRIB on several medical applications, including sleep staging on electroencephalogram (EEG) data, X-ray COVID image classification, and atrial fibrillation detection based on electrocardiogram (ECG) data. SCRIB obtained desirable class-specific risks, which are 35%-88% closer to the target risks than baseline methods.

ATD: Augmenting CP Tensor Decomposition by Self Supervision.

Tensor decompositions are powerful tools for dimensionality reduction and feature interpretation of multidimensional data such as signals. Existing tensor decomposition objectives (e.g., Frobenius norm) are designed for fitting raw data under statistical assumptions, which may not align with downstream classification tasks. In practice, raw input tensor can contain irrelevant information while data augmentation techniques may be used to smooth out class-irrelevant noise in samples. This paper addresses the above challenges by proposing augmented tensor decomposition (ATD), which effectively incorporates data augmentations and self-supervised learning (SSL) to boost downstream classification. To address the non-convexity of the new augmented objective, we develop an iterative method that enables the optimization to follow an alternating least squares (ALS) fashion. We evaluate our proposed ATD on multiple datasets. It can achieve 0.8% ~ 2.5% accuracy gain over tensor-based baselines. Also, our ATD model shows comparable or better performance (e.g., up to 15% in accuracy) over self-supervised and autoencoder baselines while using less than 5% of learnable parameters of these baseline models.

CHEER: Rich Model Helps Poor Model via Knowledge Infusion.

There is a growing interest in applying deep learning (DL) to healthcare, driven by the availability of data with multiple feature channels in rich-data environments (e.g., intensive care units). However, in many other practical situations, we can only access data with much fewer feature channels in a poor-data environments (e.g., at home), which often results in predictive models with poor performance. How can we boost the performance of models learned from such poor-data environment by leveraging knowledge extracted from existing models trained using rich data in a related environment? To address this question, we develop a knowledge infusion framework named CHEER that can succinctly summarize such rich model into transferable representations, which can be incorporated into the poor model to improve its performance. The infused model is analyzed theoretically and evaluated empirically on several datasets. Our empirical results showed that CHEER outperformed baselines by 5.60% to 46.80% in terms of the macro-F1 score on multiple physiological datasets.

MOLER: Incorporate Molecule-Level Reward to Enhance Deep Generative Model for Molecule Optimization.

The goal of molecular optimization is to generate molecules similar to a target molecule but with better chemical properties. Deep generative models have shown great success in molecule optimization. However, due to the iterative local generation process of deep generative models, the resulting molecules can significantly deviate from the input in molecular similarity and size, leading to poor chemical properties. The key issue here is that the existing deep generative models restrict their attention on substructure-level generation without considering the entire molecule as a whole. To address this challenge, we propose Molecule-Level Reward functions (MOLER) to encourage (1) the input and the generated molecule to be similar, and to ensure (2) the generated molecule has a similar size to the input. The proposed method can be combined with various deep generative models. Policy gradient technique is introduced to optimize reward-based objectives with small computational overhead. Empirical studies show that MOLER achieves up to 20.2% relative improvement in success rate over the best baseline method on several properties, including QED, DRD2 and LogP.

Machine learning applications for therapeutic tasks with genomics data.

Thanks to the increasing availability of genomics and other biomedical data, many machine learning algorithms have been proposed for a wide range of therapeutic discovery and development tasks. In this survey, we review the literature on machine learning applications for genomics through the lens of therapeutic development. We investigate the interplay among genomics, compounds, proteins, electronic health records, cellular images, and clinical texts. We identify 22 machine learning in genomics applications that span the whole therapeutics pipeline, from discovering novel targets, personalizing medicine, developing gene-editing tools, all the way to facilitating clinical trials and post-market studies. We also pinpoint seven key challenges in this field with potentials for expansion and impact. This survey examines recent research at the intersection of machine learning, genomics, and therapeutic development.

High-dynamic-range blood flow rate measurement in a large-diameter vessel.

We propose a new, to the best of our knowledge, compound technique to measure high-dynamic-range blood flow rate in a large-diameter vessel, which combines the dynamic scattering light (DLS) and the laser speckle contrast imaging (LSCI) methods, possessing the advantages of the high temporal resolution of DLS and the robust property of LSCI. By controlling the second-order spatial correlations of the laser speckle through two imaging systems, the speckle temporal intensity autocorrelation function g2(t) and the decorrelation time τc are directly measured using a high-speed camera. It turns out the enhanced spatial second-order correlation helps to measure the blood flow with higher dynamic range and that the measured parameter ß and the blood flow dynamics n were accurately determined. For further improvement the dynamic range, the modified LSCI method was adopted, and the decorrelation time as a function of blood flow rate was constructed. It reveals the feasibility of measuring the high flow rate in large-diameter vessels and provides significant guidance for the future biomedical study of the myocardial perfusion in coronary artery bypass grafting, ghost imaging, and ghost cytometry.