Reconstructing directional causal networks with random forest: Causality meeting machine learning

A methodological study of exposome based on an open database: Association analysis between exposure to metal mixtures and hyperuricemia

BACKGROUND

Exposome recognizes that humans are constantly exposed to multiple environmental factors, and elucidating the health effects of complex exposure mixtures places greater demands on analytical methods.

OBJECTS

We aimed to explore the association between mixed exposure to metals and hyperuricemia (HUA), and highlight the potential of explainable machine learning (EML) and causal mediation analysis (CMA) for application in the analysis of exposome data.

METHODS

Pre-pandemic data from the National Health and Nutrition Examination Survey (NHANES) 2011-2020 and a total of 13780 individuals were included. We first used traditional statistical models (multiple logistic regression (MLR) and restricted cubic spline regression (RCS)) and EML to explore associations between mixed metals exposures and HUA, followed by the CMA using the 4-way decomposition method to analyze the interaction and mediation effects among BMI or estimated glomerular filtration rate (eGFR), metals and HUA.

RESULTS

The prevalence of HUA was 18.91% (2606/13780). The MLR showed that mercury (Q4 vs Q1: OR = 1.08, 95% CI:1.02-1.14) and lead (Q4 vs Q1: OR = 1.23, 95% CI:1.13-1.34) were generally positively associated with HUA. Higher concentrations of lead, mercury, selenium and manganese were associated with the increased odds of HUA, and BMI and eGFR were the top two variables attributable to the risk of developing HUA in the EML. Subgroup analyses from the MLR and EML consistently demonstrated the positive relationship between exposure to lead, mercury and selenium in participants with BMI <25 kg/m2 and BMI ≥30 kg/m2. BMI mediated 32.12% of the association between lead exposure and HUA, and the interaction between BMI and lead accounted for 3.88% of the association in the CMA.

CONCLUSIONS

Heavy metals can increase the HUA risk and BMI or eGFR can mediate and interact with metals to cause HUA. Future studies based on exposome can attempt to utilize the EML and CMA.

Time series causal relationships discovery through feature importance and ensemble models

Inferring causal relationships from observational data is a key challenge in understanding the interpretability of Machine Learning models. Given the ever-increasing amount of observational data available in many areas, Machine Learning algorithms used for forecasting have become more complex, leading to a less understandable path of how a decision is made by the model. To address this issue, we propose leveraging ensemble models, e.g., Random Forest, to assess which input features the trained model prioritizes when making a forecast and, in this way, establish causal relationships between the variables. The advantage of these algorithms lies in their ability to provide feature importance, which allows us to build the causal network. We present our methodology to estimate causality in time series from oil field production. As it is difficult to extract causal relations from a real field, we also included a synthetic oil production dataset and a weather dataset, which is also synthetic, to provide the ground truth. We aim to perform causal discovery, i.e., establish the existing connections between the variables in each dataset. Through an iterative process of improving the forecasting of a target's value, we evaluate whether the forecasting improves by adding information from a new potential driver; if so, we state that the driver causally affects the target. On the oil field-related datasets, our causal analysis results agree with the interwell connections already confirmed by tracer information; whenever the tracer data are available, we used it as our ground truth. This consistency between both estimated and confirmed connections provides us the confidence about the effectiveness of our proposed methodology. To our knowledge, this is the first time causal analysis using solely production data is employed to discover interwell connections in an oil field dataset.

Conditional cross-map-based technique: From pairwise dynamical causality to causal network reconstruction

Causality detection methods based on mutual cross mapping have been fruitfully developed and applied to data originating from nonlinear dynamical systems, where the causes and effects are non-separable. However, these pairwise methods still have shortcomings in discriminating typical network structures, including common drivers, indirect dependencies, and facing the curse of dimensionality, when they are stepping to causal network reconstruction. A few endeavors have been devoted to conquer these shortcomings. Here, we propose a novel method that could be regarded as one of these endeavors. Our method, named conditional cross-map-based technique, can eliminate third-party information and successfully detect direct dynamical causality, where the detection results can exactly be categorized into four standard normal forms by the designed criterion. To demonstrate the practical usefulness of our model-free, data-driven method, data generated from different representative models covering all kinds of network motifs and measured from real-world systems are investigated. Because correct identification of the direct causal links is essential to successful modeling, predicting, and controlling the underlying complex systems, our method does shed light on uncovering the inner working mechanisms of real-world systems only using the data experimentally obtained in a variety of disciplines.

The structure inference of flocking systems based on the trajectories

The interaction between the swarm individuals affects the dynamic behavior of the swarm, but it is difficult to obtain directly from outside observation. Therefore, the problem we focus on is inferring the structure of the interactions in the swarm from the individual behavior trajectories. Similar inference problems that existed in network science are named network reconstruction or network inference. It is a fundamental problem pervading research on complex systems. In this paper, a new method, called Motion Trajectory Similarity, is developed for inferring direct interactions from the motion state of individuals in the swarm. It constructs correlations by combining the similarity of the motion trajectories of each cross section of the time series, in which individuals with highly similar motion states are more likely to interact with each other. Experiments on the flocking systems demonstrate that our method can produce a reliable interaction inference and outperform traditional network inference methods. It can withstand a high level of noise and time delay introduced into flocking models, as well as parameter variation in the flocking system, to achieve robust reconstruction. The proposed method provides a new perspective for inferring the interaction structure of a swarm, which helps us to explore the mechanisms of collective movement in swarms and paves the way for developing the flocking models that can be quantified and predicted.

Inferring causal gene regulatory network via GreyNet: From dynamic grey association to causation

Gene regulatory network (GRN) provides abundant information on gene interactions, which contributes to demonstrating pathology, predicting clinical outcomes, and identifying drug targets. Existing high-throughput experiments provide rich time-series gene expression data to reconstruct the GRN to further gain insights into the mechanism of organisms responding to external stimuli. Numerous machine-learning methods have been proposed to infer gene regulatory networks. Nevertheless, machine learning, especially deep learning, is generally a “black box,” which lacks interpretability. The causality has not been well recognized in GRN inference procedures. In this article, we introduce grey theory integrated with the adaptive sliding window technique to flexibly capture instant gene–gene interactions in the uncertain regulatory system. Then, we incorporate generalized multivariate Granger causality regression methods to transform the dynamic grey association into causation to generate directional regulatory links. We evaluate our model on the DREAM4 in silico benchmark dataset and real-world hepatocellular carcinoma (HCC) time-series data. We achieved competitive results on the DREAM4 compared with other state-of-the-art algorithms and gained meaningful GRN structure on HCC data respectively.

Quality of Life Predictors in Patients With Melanoma: A Machine Learning Approach

Health related quality of life (HRQoL) is an important recognized health outcome for cancer treatments, but also disease course with slower recovery and increased morbidity. These issues are of implication in melanoma, which maintains a risk of disease progression for many years after diagnosis. This study aimed to explore and weigh factors in the perception of the quality of life and possible relationships with demographic–clinical characteristics in people with melanoma via a machine learning approach. In this observational study, patients with melanoma, without metastatic disease, were recruited from January 2020 to December 2021 with a follow-up of at least one year. Demographic variables and clinics were collected, and the 12-Item Short-Form Health Survey (SF-12) was adopted as the physical and mental aspects of the Health-Related Quality of Life (HRQoL) measure. All the variables were processed in a random forest model to weigh at each node of each tree of this machine learning regression model, their actual weight in SF-12 score. We included 203 melanoma patients, mean aged 59.25 ± 15.1 years: 56 (27%) affecting the upper limbs and 147 (73%) affecting the trunk. The model of 142 patients with no missing value, generating 92 trees (MSE = 0.45, R2 of 0.78), reported that the lesion site was the most influencing variable on HRQoL based on the decrease in Gini impurity in variable weighing at each node intersection in forest generation. In this scenario, we built two distinct models for lesion sites and demonstrated that the variable that most influenced the quality of life in upper limb melanoma was lymphedema, while BMI was in the trunk. Given these results, random forest regressions could play a crucial role in the clinical and rehabilitation approach. The machine-learning model for detecting the HRQoL predictor in melanoma patients indicates that the experienced lymphedema and BMI may influence the HRQoL perception. This study suggests that the prevention and treatment of lymphedema and bodyweight reduction might improve the quality of life in melanoma.

Detecting interactions in discrete-time dynamics by random variable resetting

Detecting the interactions in networks helps us to understand the collective behaviors of complex systems. However, doing so is challenging due to systemic noise, nonlinearity, and a lack of information. Very few researchers have attempted to reconstruct discrete-time dynamic networks. Recently, Shi et al. proposed resetting a random state variable to infer the interactions in a continuous-time dynamic network. In this paper, we introduce a random resetting method for discrete-time dynamic networks. The statistical characteristics of the method are investigated and verified with numerical simulations. In addition, this reconstruction method was evaluated for limited data and weak coupling and within multiple-attractor systems.

Texture classification based on image (natural and horizontal) visibility graph constructing methods

Texture classification is widely used in image analysis and some other related fields. In this paper, we designed a texture classification algorithm, named by TCIVG (Texture Classification based on Image Visibility Graph), based on a newly proposed image visibility graph network constructing method by Lacasa et al. By using TCIVG on a Brodatz texture image database, the whole procedure is illustrated. First, each texture image in the image database was transformed to an associated image natural visibility graph network and an image horizontal visibility graph network. Then, the degree distribution measure [P(k)] was extracted as a key characteristic parameter to different classifiers. Numerical experiments show that for artificial texture images, a 100% classification accuracy can be obtained by means of a quadratic discriminant based on natural TCIVG. For natural texture images, 94.80% classification accuracy can be obtained by a linear SVM (Support Vector Machine) based on horizontal TCIVG. Our results are better than that reported in some existing literature studies based on the same image database.

Proposed Requirements for Cardiovascular Imaging-Related Machine Learning Evaluation (PRIME): A Checklist: Reviewed by the American College of Cardiology Healthcare Innovation Council

Machine learning (ML) has been increasingly used within cardiology, particularly in the domain of cardiovascular imaging. Due to the inherent complexity and flexibility of ML algorithms, inconsistencies in the model performance and interpretation may occur. Several review articles have been recently published that introduce the fundamental principles and clinical application of ML for cardiologists. This paper builds on these introductory principles and outlines a more comprehensive list of crucial responsibilities that need to be completed when developing ML models. This paper aims to serve as a scientific foundation to aid investigators, data scientists, authors, editors, and reviewers involved in machine learning research with the intent of uniform reporting of ML investigations. An independent multidisciplinary panel of ML experts, clinicians, and statisticians worked together to review the theoretical rationale underlying 7 sets of requirements that may reduce algorithmic errors and biases. Finally, the paper summarizes a list of reporting items as an itemized checklist that highlights steps for ensuring correct application of ML models and the consistent reporting of model specifications and results. It is expected that the rapid pace of research and development and the increased availability of real-world evidence may require periodic updates to the checklist.

Partial cross mapping eliminates indirect causal influences

Causality detection likely misidentifies indirect causations as direct ones, due to the effect of causation transitivity. Although several methods in traditional frameworks have been proposed to avoid such misinterpretations, there still is a lack of feasible methods for identifying direct causations from indirect ones in the challenging situation where the variables of the underlying dynamical system are non-separable and weakly or moderately interacting. Here, we solve this problem by developing a data-based, model-independent method of partial cross mapping based on an articulated integration of three tools from nonlinear dynamics and statistics: phase-space reconstruction, mutual cross mapping, and partial correlation. We demonstrate our method by using data from different representative models and real-world systems. As direct causations are keys to the fundamental underpinnings of a variety of complex dynamics, we anticipate our method to be indispensable in unlocking and deciphering the inner mechanisms of real systems in diverse disciplines from data.

It is crucial yet challenging to identify cause-consequence relation in complex dynamical systems where direct causal links can mix with indirect ones. Leng et al. propose a data-driven model-independent method to distinguish direct from indirect causality and test its applicability to real-world data.

Using machine learning to assess short term causal dependence and infer network links

We introduce and test a general machine-learning-based technique for the inference of short term causal dependence between state variables of an unknown dynamical system from time-series measurements of its state variables. Our technique leverages the results of a machine learning process for short time prediction to achieve our goal. The basic idea is to use the machine learning to estimate the elements of the Jacobian matrix of the dynamical flow along an orbit. The type of machine learning that we employ is reservoir computing. We present numerical tests on link inference of a network of interacting dynamical nodes. It is seen that dynamical noise can greatly enhance the effectiveness of our technique, while observational noise degrades the effectiveness. We believe that the competition between these two opposing types of noise will be the key factor determining the success of causal inference in many of the most important application situations.