Potential Predictors for Deterioration of Renal Function After Transfusion.

BACKGROUND: Transfusion of packed red blood cells (pRBCs) is still associated with risks. This study aims to determine whether renal function deterioration in the context of individual transfusions in individual patients can be predicted using machine learning. Recipient and donor characteristics linked to increased risk are identified. METHODS: This study was registered at ClinicalTrials.gov (NCT05466370) and was conducted after local ethics committee approval. We evaluated 3366 transfusion episodes from a university hospital between October 31, 2016, and August 31, 2020. Random forest models were tuned and trained via Python auto-sklearn package to predict acute kidney injury (AKI). The models included recipients' and donors' demographic parameters and laboratory values, donor questionnaire results, and the age of the pRBCs. Bootstrapping on the test dataset was used to calculate the means and standard deviations of various performance metrics. RESULTS: AKI as defined by a modified Kidney Disease Improving Global Outcomes (KDIGO) criterion developed after 17.4% transfusion episodes (base rate). AKI could be predicted with an area under the curve of the receiver operating characteristic (AUC-ROC) of 0.73 ± 0.02. The negative (NPV) and positive (PPV) predictive values were 0.90 ± 0.02 and 0.32 ± 0.03, respectively. Feature importance and relative risk analyses revealed that donor features were far less important than recipient features for predicting posttransfusion AKI. CONCLUSIONS: Surprisingly, only the recipients' characteristics played a decisive role in AKI prediction. Based on this result, we speculate that the selection of a specific pRBC may have less influence than recipient characteristics.

HyperPCM: Robust Task-Conditioned Modeling of Drug-Target Interactions.

A central problem in drug discovery is to identify the interactions between drug-like compounds and protein targets. Over the past few decades, various quantitative structure-activity relationship (QSAR) and proteo-chemometric (PCM) approaches have been developed to model and predict these interactions. While QSAR approaches solely utilize representations of the drug compound, PCM methods incorporate both representations of the protein target and the drug compound, enabling them to achieve above-chance predictive accuracy on previously unseen protein targets. Both QSAR and PCM approaches have recently been improved by machine learning and deep neural networks, that allow the development of drug-target interaction prediction models from measurement data. However, deep neural networks typically require large amounts of training data and cannot robustly adapt to new tasks, such as predicting interaction for unseen protein targets at inference time. In this work, we propose to use HyperNetworks to efficiently transfer information between tasks during inference and thus to accurately predict drug-target interactions on unseen protein targets. Our HyperPCM method reaches state-of-the-art performance compared to previous methods on multiple well-known benchmarks, including Davis, DUD-E, and a ChEMBL derived data set, and particularly excels at zero-shot inference involving unseen protein targets. Our method, as well as reproducible data preparation, is available at https://github.com/ml-jku/hyper-dti.

A community effort in SARS-CoV-2 drug discovery.

The COVID-19 pandemic continues to pose a substantial threat to human lives and is likely to do so for years to come. Despite the availability of vaccines, searching for efficient small-molecule drugs that are widely available, including in low- and middle-income countries, is an ongoing challenge. In this work, we report the results of an open science community effort, the "Billion molecules against COVID-19 challenge", to identify small-molecule inhibitors against SARS-CoV-2 or relevant human receptors. Participating teams used a wide variety of computational methods to screen a minimum of 1 billion virtual molecules against 6 protein targets. Overall, 31 teams participated, and they suggested a total of 639,024 molecules, which were subsequently ranked to find 'consensus compounds'. The organizing team coordinated with various contract research organizations (CROs) and collaborating institutions to synthesize and test 878 compounds for biological activity against proteases (Nsp5, Nsp3, TMPRSS2), nucleocapsid N, RdRP (only the Nsp12 domain), and (alpha) spike protein S. Overall, 27 compounds with weak inhibition/binding were experimentally identified by binding-, cleavage-, and/or viral suppression assays and are presented here. Open science approaches such as the one presented here contribute to the knowledge base of future drug discovery efforts in finding better SARS-CoV-2 treatments.

CLOOME: contrastive learning unlocks bioimaging databases for queries with chemical structures.

The field of bioimage analysis is currently impacted by a profound transformation, driven by the advancements in imaging technologies and artificial intelligence. The emergence of multi-modal AI systems could allow extracting and utilizing knowledge from bioimaging databases based on information from other data modalities. We leverage the multi-modal contrastive learning paradigm, which enables the embedding of both bioimages and chemical structures into a unified space by means of bioimage and molecular structure encoders. This common embedding space unlocks the possibility of querying bioimaging databases with chemical structures that induce different phenotypic effects. Concretely, in this work we show that a retrieval system based on multi-modal contrastive learning is capable of identifying the correct bioimage corresponding to a given chemical structure from a database of ~2000 candidate images with a top-1 accuracy >70 times higher than a random baseline. Additionally, the bioimage encoder demonstrates remarkable transferability to various further prediction tasks within the domain of drug discovery, such as activity prediction, molecule classification, and mechanism of action identification. Thus, our approach not only addresses the current limitations of bioimaging databases but also paves the way towards foundation models for microscopy images.

Txt2Img-MHN: Remote Sensing Image Generation From Text Using Modern Hopfield Networks.

The synthesis of high-resolution remote sensing images based on text descriptions has great potential in many practical application scenarios. Although deep neural networks have achieved great success in many important remote sensing tasks, generating realistic remote sensing images from text descriptions is still very difficult. To address this challenge, we propose a novel text-to-image modern Hopfield network (Txt2Img-MHN). The main idea of Txt2Img-MHN is to conduct hierarchical prototype learning on both text and image embeddings with modern Hopfield layers. Instead of directly learning concrete but highly diverse text-image joint feature representations for different semantics, Txt2Img-MHN aims to learn the most representative prototypes from text-image embeddings, achieving a coarse-to-fine learning strategy. These learned prototypes can then be utilized to represent more complex semantics in the text-to-image generation task. To better evaluate the realism and semantic consistency of the generated images, we further conduct zero-shot classification on real remote sensing data using the classification model trained on synthesized images. Despite its simplicity, we find that the overall accuracy in the zero-shot classification may serve as a good metric to evaluate the ability to generate an image from text. Extensive experiments on the benchmark remote sensing text-image dataset demonstrate that the proposed Txt2Img-MHN can generate more realistic remote sensing images than existing methods. Code and pre-trained models are available online (https://github.com/YonghaoXu/Txt2Img-MHN).

Using emergency department triage for machine learning-based admission and mortality prediction.

AIMS: Patient admission is a decision relying on sparsely available data. This study aims to provide prediction models for discharge versus admission for ward observation or intensive care, and 30 day-mortality for patients triaged with the Manchester Triage System. METHODS: This is a single-centre, observational, retrospective cohort study from data within ten minutes of patient presentation at the interdisciplinary emergency department of the Kepler University Hospital, Linz, Austria. We trained machine learning models including Random Forests and Neural Networks individually to predict discharge versus ward observation or intensive care admission, and 30 day-mortality. For analysis of the features' relevance, we used permutation feature importance. RESULTS: A total of 58323 adult patients between 1 December 2015 and 31 August 2020 were included. Neural Networks and Random Forests predicted admission to ward observation with an AUC-ROC of 0.842 ±â€…0.00 with the most important features being age and chief complaint. For admission to intensive care, the models had an AUC-ROC of 0.819 ±â€…0.002 with the most important features being the Manchester Triage category and heart rate, and for the outcome 30 day-mortality an AUC-ROC of 0.925 ±â€…0.001. The most important features for the prediction of 30 day-mortality were age and general ward admission. CONCLUSION: Machine learning can provide prediction on discharge versus admission to general wards and intensive care and inform about risk on 30 day-mortality for patients in the emergency department.

In Defense of Metrics: Metrics Sufficiently Encode Typical Human Preferences Regarding Hydrological Model Performance.

Building accurate rainfall-runoff models is an integral part of hydrological science and practice. The variety of modeling goals and applications have led to a large suite of evaluation metrics for these models. Yet, hydrologists still put considerable trust into visual judgment, although it is unclear whether such judgment agrees or disagrees with existing quantitative metrics. In this study, we tasked 622 experts to compare and judge more than 14,000 pairs of hydrographs from 13 different models. Our results show that expert opinion broadly agrees with quantitative metrics and results in a clear preference for a Machine Learning model over traditional hydrological models. The expert opinions are, however, subject to significant amounts of inconsistency. Nevertheless, where experts agree, we can predict their opinion purely from quantitative metrics, which indicates that the metrics sufficiently encode human preferences in a small set of numbers. While there remains room for improvement of quantitative metrics, we suggest that the hydrologic community should reinforce their benchmarking efforts and put more trust in these metrics.

Lifting Hospital Electronic Health Record Data Treasures: Challenges and Opportunities.

Electronic health records (EHRs) have been successfully used in data science and machine learning projects. However, most of these data are collected for clinical use rather than for retrospective analysis. This means that researchers typically face many different issues when attempting to access and prepare the data for secondary use. We aimed to investigate how raw EHRs can be accessed and prepared in retrospective data science projects in a disciplined, effective, and efficient way. We report our experience and findings from a large-scale data science project analyzing routinely acquired retrospective data from the Kepler University Hospital in Linz, Austria. The project involved data collection from more than 150,000 patients over a period of 10 years. It included diverse data modalities, such as static demographic data, irregularly acquired laboratory test results, regularly sampled vital signs, and high-frequency physiological waveform signals. Raw medical data can be corrupted in many unexpected ways that demand thorough manual inspection and highly individualized data cleaning solutions. We present a general data preparation workflow, which was shaped in the course of our project and consists of the following 7 steps: obtain a rough overview of the available EHR data, define clinically meaningful labels for supervised learning, extract relevant data from the hospital's data warehouses, match data extracted from different sources, deidentify them, detect errors and inconsistencies therein through a careful exploratory analysis, and implement a suitable data processing pipeline in actual code. Only few of the data preparation issues encountered in our project were addressed by generic medical data preprocessing tools that have been proposed recently. Instead, highly individualized solutions for the specific data used in one's own research seem inevitable. We believe that the proposed workflow can serve as a guidance for practitioners, helping them to identify and address potential problems early and avoid some common pitfalls.

In silico proof of principle of machine learning-based antibody design at unconstrained scale.

Generative machine learning (ML) has been postulated to become a major driver in the computational design of antigen-specific monoclonal antibodies (mAb). However, efforts to confirm this hypothesis have been hindered by the infeasibility of testing arbitrarily large numbers of antibody sequences for their most critical design parameters: paratope, epitope, affinity, and developability. To address this challenge, we leveraged a lattice-based antibody-antigen binding simulation framework, which incorporates a wide range of physiological antibody-binding parameters. The simulation framework enables the computation of synthetic antibody-antigen 3D-structures, and it functions as an oracle for unrestricted prospective evaluation and benchmarking of antibody design parameters of ML-generated antibody sequences. We found that a deep generative model, trained exclusively on antibody sequence (one dimensional: 1D) data can be used to design conformational (three dimensional: 3D) epitope-specific antibodies, matching, or exceeding the training dataset in affinity and developability parameter value variety. Furthermore, we established a lower threshold of sequence diversity necessary for high-accuracy generative antibody ML and demonstrated that this lower threshold also holds on experimental real-world data. Finally, we show that transfer learning enables the generation of high-affinity antibody sequences from low-N training data. Our work establishes a priori feasibility and the theoretical foundation of high-throughput ML-based mAb design.

Domain Shifts in Machine Learning Based Covid-19 Diagnosis From Blood Tests.

Many previous studies claim to have developed machine learning models that diagnose COVID-19 from blood tests. However, we hypothesize that changes in the underlying distribution of the data, so called domain shifts, affect the predictive performance and reliability and are a reason for the failure of such machine learning models in clinical application. Domain shifts can be caused, e.g., by changes in the disease prevalence (spreading or tested population), by refined RT-PCR testing procedures (way of taking samples, laboratory procedures), or by virus mutations. Therefore, machine learning models for diagnosing COVID-19 or other diseases may not be reliable and degrade in performance over time. We investigate whether domain shifts are present in COVID-19 datasets and how they affect machine learning methods. We further set out to estimate the mortality risk based on routinely acquired blood tests in a hospital setting throughout pandemics and under domain shifts. We reveal domain shifts by evaluating the models on a large-scale dataset with different assessment strategies, such as temporal validation. We present the novel finding that domain shifts strongly affect machine learning models for COVID-19 diagnosis and deteriorate their predictive performance and credibility. Therefore, frequent re-training and re-assessment are indispensable for robust models enabling clinical utility.

Improving Few- and Zero-Shot Reaction Template Prediction Using Modern Hopfield Networks.

Finding synthesis routes for molecules of interest is essential in the discovery of new drugs and materials. To find such routes, computer-assisted synthesis planning (CASP) methods are employed, which rely on a single-step model of chemical reactivity. In this study, we introduce a template-based single-step retrosynthesis model based on Modern Hopfield Networks, which learn an encoding of both molecules and reaction templates in order to predict the relevance of templates for a given molecule. The template representation allows generalization across different reactions and significantly improves the performance of template relevance prediction, especially for templates with few or zero training examples. With inference speed up to orders of magnitude faster than baseline methods, we improve or match the state-of-the-art performance for top-k exact match accuracy for k ≥ 3 in the retrosynthesis benchmark USPTO-50k. Code to reproduce the results is available at github.com/ml-jku/mhn-react.

Machine Learning-Based Mortality Prediction of Patients at Risk During Hospital Admission.

OBJECTIVES: The ability to predict in-hospital mortality from data available at hospital admission would identify patients at risk and thereby assist hospital-wide patient safety initiatives. Our aim was to use modern machine learning tools to predict in-hospital mortality from standardized data sets available at hospital admission. METHODS: This was a retrospective, observational study in 3 adult tertiary care hospitals in Western Australia between January 2008 and June 2017. Primary outcome measures were the area under the curve for the receiver operating characteristics curve, the F1 score, and the average precision of the 4 machine learning algorithms used: logistic regression, neural networks, random forests, and gradient boosting trees. RESULTS: Using our 4 predictive models, in-hospital mortality could be predicted satisfactorily (areas under the curve for neural networks, logistic regression, random forests, and gradient boosting trees: 0.932, 0.936, 0.935, and 0.935, respectively), with moderate F1 scores: 0.378, 0.367, 0.380, and 0.380, respectively. Average precision values were 0.312, 0.321, 0.334, and 0.323, respectively. It remains unknown whether additional features might improve our models; however, this would result in additional efforts for data acquisition in daily clinical practice. CONCLUSIONS: This study demonstrates that using only a limited, standardized data set in-hospital mortality can be predicted satisfactorily at the time point of hospital admission. More parameters describing patient's health are likely needed to improve our model.

Unconstrained generation of synthetic antibody-antigen structures to guide machine learning methodology for antibody specificity prediction.

Machine learning (ML) is a key technology for accurate prediction of antibody-antigen binding. Two orthogonal problems hinder the application of ML to antibody-specificity prediction and the benchmarking thereof: the lack of a unified ML formalization of immunological antibody-specificity prediction problems and the unavailability of large-scale synthetic datasets to benchmark real-world relevant ML methods and dataset design. Here we developed the Absolut! software suite that enables parameter-based unconstrained generation of synthetic lattice-based three-dimensional antibody-antigen-binding structures with ground-truth access to conformational paratope, epitope and affinity. We formalized common immunological antibody-specificity prediction problems as ML tasks and confirmed that for both sequence- and structure-based tasks, accuracy-based rankings of ML methods trained on experimental data hold for ML methods trained on Absolut!-generated data. The Absolut! framework has the potential to enable real-world relevant development and benchmarking of ML strategies for biotherapeutics design.

The Promise of AI for DILI Prediction.

Drug-induced liver injury (DILI) is a common reason for the withdrawal of a drug from the market. Early assessment of DILI risk is an essential part of drug development, but it is rendered challenging prior to clinical trials by the complex factors that give rise to liver damage. Artificial intelligence (AI) approaches, particularly those building on machine learning, range from random forests to more recent techniques such as deep learning, and provide tools that can analyze chemical compounds and accurately predict some of their properties based purely on their structure. This article reviews existing AI approaches to predicting DILI and elaborates on the challenges that arise from the as yet limited availability of data. Future directions are discussed focusing on rich data modalities, such as 3D spheroids, and the slow but steady increase in drugs annotated with DILI risk labels.

Artificial neural networks and pathologists recognize basal cell carcinomas based on different histological patterns.

Recent advances in artificial intelligence, particularly in the field of deep learning, have enabled researchers to create compelling algorithms for medical image analysis. Histological slides of basal cell carcinomas (BCCs), the most frequent skin tumor, are accessed by pathologists on a daily basis and are therefore well suited for automated prescreening by neural networks for the identification of cancerous regions and swift tumor classification.In this proof-of-concept study, we implemented an accurate and intuitively interpretable artificial neural network (ANN) for the detection of BCCs in histological whole-slide images (WSIs). Furthermore, we identified and compared differences in the diagnostic histological features and recognition patterns relevant for machine learning algorithms vs. expert pathologists.An attention-ANN was trained with WSIs of BCCs to identify tumor regions (n = 820). The diagnosis-relevant regions used by the ANN were compared to regions of interest for pathologists, detected by eye-tracking techniques.This ANN accurately identified BCC tumor regions on images of histologic slides (area under the ROC curve: 0.993, 95% CI: 0.990-0.995; sensitivity: 0.965, 95% CI: 0.951-0.979; specificity: 0.910, 95% CI: 0.859-0.960). The ANN implicitly calculated a weight matrix, indicating the regions of a histological image that are important for the prediction of the network. Interestingly, compared to pathologists' eye-tracking results, machine learning algorithms rely on significantly different recognition patterns for tumor identification (p < 10-4).To conclude, we found on the example of BCC WSIs, that histopathological images can be efficiently and interpretably analyzed by state-of-the-art machine learning techniques. Neural networks and machine learning algorithms can potentially enhance diagnostic precision in digital pathology and uncover hitherto unused classification patterns.

Machine learning-based prediction of transfusion.

BACKGROUND: The ability to predict transfusions arising during hospital admission might enable economized blood supply management and might furthermore increase patient safety by ensuring a sufficient stock of red blood cells (RBCs) for a specific patient. We therefore investigated the precision of four different machine learning-based prediction algorithms to predict transfusion, massive transfusion, and the number of transfusions in patients admitted to a hospital. STUDY DESIGN AND METHODS: This was a retrospective, observational study in three adult tertiary care hospitals in Western Australia between January 2008 and June 2017. Primary outcome measures for the classification tasks were the area under the curve for the receiver operating characteristics curve, the F1 score, and the average precision of the four machine learning algorithms used: neural networks (NNs), logistic regression (LR), random forests (RFs), and gradient boosting (GB) trees. RESULTS: Using our four predictive models, transfusion of at least 1 unit of RBCs could be predicted rather accurately (sensitivity for NN, LR, RF, and GB: 0.898, 0.894, 0.584, and 0.872, respectively; specificity: 0.958, 0.966, 0.964, 0.965). Using the four methods for prediction of massive transfusion was less successful (sensitivity for NN, LR, RF, and GB: 0.780, 0.721, 0.002, and 0.797, respectively; specificity: 0.994, 0.995, 0.993, 0.995). As a consequence, prediction of the total number of packed RBCs transfused was also rather inaccurate. CONCLUSION: This study demonstrates that the necessity for intrahospital transfusion can be forecasted reliably, however the amount of RBC units transfused during a hospital stay is more difficult to predict.

Industry-scale application and evaluation of deep learning for drug target prediction.

Artificial intelligence (AI) is undergoing a revolution thanks to the breakthroughs of machine learning algorithms in computer vision, speech recognition, natural language processing and generative modelling. Recent works on publicly available pharmaceutical data showed that AI methods are highly promising for Drug Target prediction. However, the quality of public data might be different than that of industry data due to different labs reporting measurements, different measurement techniques, fewer samples and less diverse and specialized assays. As part of a European funded project (ExCAPE), that brought together expertise from pharmaceutical industry, machine learning, and high-performance computing, we investigated how well machine learning models obtained from public data can be transferred to internal pharmaceutical industry data. Our results show that machine learning models trained on public data can indeed maintain their predictive power to a large degree when applied to industry data. Moreover, we observed that deep learning derived machine learning models outperformed comparable models, which were trained by other machine learning algorithms, when applied to internal pharmaceutical company datasets. To our knowledge, this is the first large-scale study evaluating the potential of machine learning and especially deep learning directly at the level of industry-scale settings and moreover investigating the transferability of publicly learned target prediction models towards industrial bioactivity prediction pipelines.

Accurate Prediction of Biological Assays with High-Throughput Microscopy Images and Convolutional Networks.

Predicting the outcome of biological assays based on high-throughput imaging data is a highly promising task in drug discovery since it can tremendously increase hit rates and suggest novel chemical scaffolds. However, end-to-end learning with convolutional neural networks (CNNs) has not been assessed for the task biological assay prediction despite the success of these networks at visual recognition. We compared several CNNs trained directly on high-throughput imaging data to a) CNNs trained on cell-centric crops and to b) the current state-of-the-art: fully connected networks trained on precalculated morphological cell features. The comparison was performed on the Cell Painting data set, the largest publicly available data set of microscopic images of cells with approximately 30,000 compound treatments. We found that CNNs perform significantly better at predicting the outcome of assays than fully connected networks operating on precomputed morphological features of cells. Surprisingly, the best performing method could predict 32% of the 209 biological assays at high predictive performance (AUC > 0.9) indicating that the cell morphology changes contain a large amount of information about compound activities. Our results suggest that many biological assays could be replaced by high-throughput imaging together with convolutional neural networks and that the costly cell segmentation and feature extraction step can be replaced by convolutional neural networks.

On failure modes in molecule generation and optimization.

There has been a wave of generative models for molecules triggered by advances in the field of Deep Learning. These generative models are often used to optimize chemical compounds towards particular properties or a desired biological activity. The evaluation of generative models remains challenging and suggested performance metrics or scoring functions often do not cover all relevant aspects of drug design projects. In this work, we highlight some unintended failure modes in molecular generation and optimization and how these evade detection by current performance metrics.