Machine Learning for Causal Inference in Biological Networks: Perspectives of This Challenge

Abstract

Most machine learning-based methods predict outcomes rather than understanding causality. Machine learning methods have been proved to be efficient in finding correlations in data, but unskilful to determine causation. This issue severely limits the applicability of machine learning methods to infer the causal relationships between the entities of a biological network, and more in general of any dynamical system, such as medical intervention strategies and clinical outcomes system, that is representable as a network. From the perspective of those who want to use the results of network inference not only to understand the mechanisms underlying the dynamics, but also to understand how the network reacts to external stimuli (e. g. environmental factors, therapeutic treatments), tools that can understand the causal relationships between data are highly demanded. Given the increasing popularity of machine learning techniques in computational biology and the recent literature proposing the use of machine learning techniques for the inference of biological networks, we would like to present the challenges that mathematics and computer science research faces in generalising machine learning to an approach capable of understanding causal relationships, and the prospects that achieving this will open up for the medical application domains of systems biology, the main paradigm of which is precisely network biology at any physical scale.

An introduction to causal inference for pharmacometricians

“…The use of DAGs for covariate selection in epidemiological studies of relevance to pharmacometrics, for example, for characterizing longitudinal progression toward end‐stage renal disease 10 or for characterizing overall survival in oncology in response to immune checkpoint inhibitors (CPIs) 11,12 The application of interpretable artificial intelligence/machine‐learning (AI/ML) algorithms (e.g., with interpretation assisted by Shapley values) to population pharmacokinetic modeling 13 and prediction of relapse and related disease activity in multiple sclerosis, 14 contemporaneous with an increased recognition of the interpretive value of formal causal frameworks in AI/ML research 15–18 The advent of real‐world evidence (RWE) usage in pharmacometric analyses, 19 contemporaneous with a growing body of guidance for the use of RWE that advocates for the use of causal DAGs 20 …”

“… 11 , 12 The application of interpretable artificial intelligence/machine‐learning (AI/ML) algorithms (e.g., with interpretation assisted by Shapley values) to population pharmacokinetic modeling 13 and prediction of relapse and related disease activity in multiple sclerosis, 14 contemporaneous with an increased recognition of the interpretive value of formal causal frameworks in AI/ML research. 15 , 16 , 17 , 18 The advent of real‐world evidence (RWE) usage in pharmacometric analyses, 19 contemporaneous with a growing body of guidance for the use of RWE that advocates for the use of causal DAGs. 20 The increasingly favorable environment for employing external or synthetic control arms in clinical trials, with the intent of generating estimates of (causal) treatment effects using methods that approximate the effects of randomization.…”

“…The application of interpretable artificial intelligence/machine‐learning (AI/ML) algorithms (e.g., with interpretation assisted by Shapley values) to population pharmacokinetic modeling 13 and prediction of relapse and related disease activity in multiple sclerosis, 14 contemporaneous with an increased recognition of the interpretive value of formal causal frameworks in AI/ML research. 15 , 16 , 17 , 18…”

CaRTS: Causality-driven Robot Tool Segmentation from Vision and Kinematics Data

“…Lenis et al [17] use the concept of counterfactuals to analyze and interpret medical image classifiers. Lecca [16] provides perspectives on the challenges of machine learning algorithms for causal inference in biological networks. Further, causality-inspired methods [35,21,23,34,18] exist for domain generalization [33].…”

Unified Tumor Growth Mechanisms from Multimodel Inference and Dataset Integration

“…To address this challenge, methods such as model selection and multimodel inference have been proposed using information theoretic scoring techniques such as Akaike Information Criterion, (AIC) with limited success stemming from difficulties with model averaging (11,12) and the fact that AIC scores do not inherently describe whether a model or features within are informed by the data. More recently, the use of AI and machine learning approaches has given impetus to causal relationship inference (13) but these relationships remain difficult to elucidate, thus underscoring the need for both novel tools for hypothesis exploration, and tools that can be used with rigor in the face of missing data that may not inform all hypotheses.…”