A spectrum of explainable and interpretable machine learning approaches for genomic studies

Abstract

The advancement of high‐throughput genomic assays has led to enormous growth in the availability of large‐scale biological datasets. Over the last two decades, these increasingly complex data have required statistical approaches that are more sophisticated than traditional linear models. Machine learning methodologies such as neural networks have yielded state‐of‐the‐art performance for prediction‐based tasks in many biomedical applications. However, a notable downside of these machine learning models is that they typically do not reveal how or why accurate predictions are made. In many areas of biomedicine, this “black box” property can be less than desirable—particularly when there is a need to perform in silico hypothesis testing about a biological system, in addition to justifying model findings for downstream decision‐making, such as determining the best next experiment or treatment strategy. Explainable and interpretable machine learning approaches have emerged to overcome this issue. While explainable methods attempt to derive post hoc understanding of what a model has learned, interpretable models are designed to inherently provide an intelligible definition of their parameters and architecture. Here, we review the model transparency spectrum moving from black box and explainable, to interpretable machine learning methodology. Motivated by applications in genomics, we provide background on the advances across this spectrum, detailing specific approaches in both supervised and unsupervised learning. Importantly, we focus on the promise of incorporating existing biological knowledge when constructing interpretable machine learning methods for biomedical applications. We then close with considerations and opportunities for new development in this space.This article is categorized under: Statistical Models > Nonlinear Models Applications of Computational Statistics > Genomics/Proteomics/Genetics Applications of Computational Statistics > Computational and Molecular Biology

Progress of Crude Oil Gasification Technology Assisted by Microorganisms in Reservoirs

“…These genera have been recognized for their significant role in petroleum hydrocarbon degradation. The isolated fermentation bacteria primarily belong to the genera Thermococcus, Bacteroidetes, Acinetobacter, and Haloanaerobium, among others [15][16][17]. The majority of reduced strains are classified within the genera Smithella, Desulfovibrio, and Desulfobulbus.…”

“…To maximize the efficiency of crude oil gasification, Lin et al [13] systematically researched the impact of temperature, pH, carbon sources, and other factors, elucidating the straightforward correlations between individual variables and the efficiency of crude oil gasification. As artificial intelligence has advanced, machine learning has demonstrated distinct advantages in fitting predictions and evaluating the significance of multiple variables [14,15]. Building on prior research regarding the correlation between individual variables and the efficiency of crude oil gasification, the integration of machine-learning algorithms is anticipated to enable the evaluation of the significance of multiple variables in the crude oil gasification process [16], thus aiding in the precise control of crude oil gasification in engineering applications.…”

Generative modeling of biological shapes and images using a probabilisticα-shape sampler

“…For example, to sample random points from a probability distribution over a manifold, one theoretically needs to know the manifold itself which can be impractical to estimate for many applications [21][22][23][24] . Recently, there are have been machine learning algorithms that have been developed for generating point clouds and reconstructing shapes using dual generators 25 , diffusion-based methods 26 , encoders 27 , and generative adversarial networks 28,29 ; but each of these frameworks lack transparency into the generative process for creating new synthetic shapes 30 . From a more mathematical perspective, several methods have been proposed to infer shapes from randomly generated point clouds.…”

Interpretable machine learning decodes soil microbiome’s response to drought stress

“…ML models often resemble black boxes, with their internal decision-making obscured from users. The interpretation of the reasons for certain predictions is essential, especially for complex biological data [17][18][19]. This is where interpretable ML methods such as SHapley Additive ExPlanation (SHAP) values are applied [20].…”