ExplaiNAble BioLogical Age (ENABL Age): an artificial intelligence framework for interpretable biological age.

BACKGROUND: Biological age is a measure of health that offers insights into ageing. The existing age clocks, although valuable, often trade off accuracy and interpretability. We introduce ExplaiNAble BioLogical Age (ENABL Age), a computational framework that combines machine-learning models with explainable artificial intelligence (XAI) methods to accurately estimate biological age with individualised explanations. METHODS: To construct the ENABL Age clock, we first predicted an age-related outcome (eg, all-cause or cause-specific mortality), and then rescaled these predictions to estimate biological age, using UK Biobank and National Health and Nutrition Examination Survey (NHANES) datasets. We adapted existing XAI methods to decompose individual ENABL Ages into contributing risk factors. For broad accessibility, we developed two versions: ENABL Age-L, based on blood tests, and ENABL Age-Q, based on questionnaire characteristics. Finally, we validated diverse ageing mechanisms captured by each ENABL Age clock through genome-wide association studies (GWAS) association analyses. FINDINGS: Our ENABL Age clock was significantly correlated with chronological age (r=0·7867, p<0·0001 for UK Biobank; r=0·7126, p<0·0001 for NHANES). These clocks distinguish individuals who are healthy (ie, their ENABL Age is lower than their chronological age) from those who are unhealthy (ie, their ENABL Age is higher than their chronological age), predicting mortality more effectively than existing clocks. Groups of individuals who were unhealthy showed approximately three to 12 times higher log hazard ratio than healthy groups, as per ENABL Age. The clocks achieved high mortality prediction power with an area under the receiver operating characteristic curve of 0·8179 for 5-year mortality and 0·8115 for 10-year mortality on the UK Biobank dataset, and 0·8935 for 5-year mortality and 0·9107 for 10-year mortality on the NHANES dataset. The individualised explanations that revealed the contribution of specific characteristics to ENABL Age provided insights into the important characteristics for ageing. An association analysis with risk factors and ageing-related morbidities and GWAS results on ENABL Age clocks trained on different mortality causes showed that each clock captures distinct ageing mechanisms. INTERPRETATION: ENABL Age brings an important leap forward in the application of XAI for interpreting biological age clocks. ENABL Age also carries substantial potential in practical settings, assisting medical professionals in untangling the complexity of ageing mechanisms, and potentially becoming a valuable tool in informed clinical decision-making processes. FUNDING: National Science Foundation and National Institutes of Health.

Uncovering expression signatures of synergistic drug responses via ensembles of explainable machine-learning models.

Machine learning may aid the choice of optimal combinations of anticancer drugs by explaining the molecular basis of their synergy. By combining accurate models with interpretable insights, explainable machine learning promises to accelerate data-driven cancer pharmacology. However, owing to the highly correlated and high-dimensional nature of transcriptomic data, naively applying current explainable machine-learning strategies to large transcriptomic datasets leads to suboptimal outcomes. Here by using feature attribution methods, we show that the quality of the explanations can be increased by leveraging ensembles of explainable machine-learning models. We applied the approach to a dataset of 133 combinations of 46 anticancer drugs tested in ex vivo tumour samples from 285 patients with acute myeloid leukaemia and uncovered a haematopoietic-differentiation signature underlying drug combinations with therapeutic synergy. Ensembles of machine-learning models trained to predict drug combination synergies on the basis of gene-expression data may improve the feature attribution quality of complex machine-learning models.

Interpretable machine learning prediction of all-cause mortality.

Background: Unlike linear models which are traditionally used to study all-cause mortality, complex machine learning models can capture non-linear interrelations and provide opportunities to identify unexplored risk factors. Explainable artificial intelligence can improve prediction accuracy over linear models and reveal great insights into outcomes like mortality. This paper comprehensively analyzes all-cause mortality by explaining complex machine learning models. Methods: We propose the IMPACT framework that uses XAI technique to explain a state-of-the-art tree ensemble mortality prediction model. We apply IMPACT to understand all-cause mortality for 1-, 3-, 5-, and 10-year follow-up times within the NHANES dataset, which contains 47,261 samples and 151 features. Results: We show that IMPACT models achieve higher accuracy than linear models and neural networks. Using IMPACT, we identify several overlooked risk factors and interaction effects. Furthermore, we identify relationships between laboratory features and mortality that may suggest adjusting established reference intervals. Finally, we develop highly accurate, efficient and interpretable mortality risk scores that can be used by medical professionals and individuals without medical expertise. We ensure generalizability by performing temporal validation of the mortality risk scores and external validation of important findings with the UK Biobank dataset. Conclusions: IMPACT's unique strength is the explainable prediction, which provides insights into the complex, non-linear relationships between mortality and features, while maintaining high accuracy. Our explainable risk scores could help individuals improve self-awareness of their health status and help clinicians identify patients with high risk. IMPACT takes a consequential step towards bringing contemporary developments in XAI to epidemiology.

This study identifies characteristics that will make a person more likely to die sooner than expected based on life expectancy for the population. We developed a computer program and applied it to information obtained about the characteristics and medical history of people from the USA. We identified previously unidentified characteristics that impact how likely it is someone will die sooner than expected, for example the circumference of the arm. We also identified combinations of characteristics that interact to increase the likelihood of death sooner than expected. By adding a person's characteristics to the program, the likelihood of death over the next 5 years can be calculated and characteristics identified that a person could modify to improve their health and reduce their chance of death during this period.

Explaining a series of models by propagating Shapley values.

Local feature attribution methods are increasingly used to explain complex machine learning models. However, current methods are limited because they are extremely expensive to compute or are not capable of explaining a distributed series of models where each model is owned by a separate institution. The latter is particularly important because it often arises in finance where explanations are mandated. Here, we present Generalized DeepSHAP (G-DeepSHAP), a tractable method to propagate local feature attributions through complex series of models based on a connection to the Shapley value. We evaluate G-DeepSHAP across biological, health, and financial datasets to show that it provides equally salient explanations an order of magnitude faster than existing model-agnostic attribution techniques and demonstrate its use in an important distributed series of models setting.

Forecasting adverse surgical events using self-supervised transfer learning for physiological signals.

Hundreds of millions of surgical procedures take place annually across the world, which generate a prevalent type of electronic health record (EHR) data comprising time series physiological signals. Here, we present a transferable embedding method (i.e., a method to transform time series signals into input features for predictive machine learning models) named PHASE (PHysiologicAl Signal Embeddings) that enables us to more accurately forecast adverse surgical outcomes based on physiological signals. We evaluate PHASE on minute-by-minute EHR data of more than 50,000 surgeries from two operating room (OR) datasets and patient stays in an intensive care unit (ICU) dataset. PHASE outperforms other state-of-the-art approaches, such as long-short term memory networks trained on raw data and gradient boosted trees trained on handcrafted features, in predicting six distinct outcomes: hypoxemia, hypocapnia, hypotension, hypertension, phenylephrine, and epinephrine. In a transfer learning setting where we train embedding models in one dataset then embed signals and predict adverse events in unseen data, PHASE achieves significantly higher prediction accuracy at lower computational cost compared to conventional approaches. Finally, given the importance of understanding models in clinical applications we demonstrate that PHASE is explainable and validate our predictive models using local feature attribution methods.

From Local Explanations to Global Understanding with Explainable AI for Trees.

Tree-based machine learning models such as random forests, decision trees, and gradient boosted trees are popular non-linear predictive models, yet comparatively little attention has been paid to explaining their predictions. Here, we improve the interpretability of tree-based models through three main contributions: 1) The first polynomial time algorithm to compute optimal explanations based on game theory. 2) A new type of explanation that directly measures local feature interaction effects. 3) A new set of tools for understanding global model structure based on combining many local explanations of each prediction. We apply these tools to three medical machine learning problems and show how combining many high-quality local explanations allows us to represent global structure while retaining local faithfulness to the original model. These tools enable us to i) identify high magnitude but low frequency non-linear mortality risk factors in the US population, ii) highlight distinct population sub-groups with shared risk characteristics, iii) identify non-linear interaction effects among risk factors for chronic kidney disease, and iv) monitor a machine learning model deployed in a hospital by identifying which features are degrading the model's performance over time. Given the popularity of tree-based machine learning models, these improvements to their interpretability have implications across a broad set of domains.

Probabilistic model-based approach for heart beat detection.

Nowadays, hospitals are ubiquitous and integral to modern society. Patients flow in and out of a veritable whirlwind of paperwork, consultations, and potential inpatient admissions, through an abstracted system that is not without flaws. One of the biggest flaws in the medical system is perhaps an unexpected one: the patient alarm system. One longitudinal study reported an 88.8% rate of false alarms, with other studies reporting numbers of similar magnitudes. These false alarm rates lead to deleterious effects that manifest in a lower standard of care across clinics. This paper discusses a model-based probabilistic inference approach to estimate physiological variables at a detection level. We design a generative model that complies with a layman's understanding of human physiology and perform approximate Bayesian inference. One primary goal of this paper is to justify a Bayesian modeling approach to increasing robustness in a physiological domain. In order to evaluate our algorithm we look at the application of heart beat detection using four datasets provided by PhysioNet, a research resource for complex physiological signals, in the form of the PhysioNet 2014 Challenge set-p1 and set-p2, the MIT-BIH Polysomnographic Database, and the MGH/MF Waveform Database. On these data sets our algorithm performs on par with the other top six submissions to the PhysioNet 2014 challenge. The overall evaluation scores in terms of sensitivity and positive predictivity values obtained were as follows: set-p1 (99.72%), set-p2 (93.51%), MIT-BIH (99.66%), and MGH/MF (95.53%). These scores are based on the averaging of gross sensitivity, gross positive predictivity, average sensitivity, and average positive predictivity.