Explainable AI for Bioinformatics: Methods, Tools and Applications

Abstract

Artificial intelligence (AI) systems utilizing deep neural networks and machine learning (ML) algorithms are widely used for solving critical problems in bioinformatics, biomedical informatics and precision medicine. However, complex ML models that are often perceived as opaque and black-box methods make it difficult to understand the reasoning behind their decisions. This lack of transparency can be a challenge for both end-users and decision-makers, as well as AI developers. In sensitive areas such as healthcare, explainability and accountability are not only desirable properties but also legally required for AI systems that can have a significant impact on human lives. Fairness is another growing concern, as algorithmic decisions should not show bias or discrimination towards certain groups or individuals based on sensitive attributes. Explainable AI (XAI) aims to overcome the opaqueness of black-box models and to provide transparency in how AI systems make decisions. Interpretable ML models can explain how they make predictions and identify factors that influence their outcomes. However, the majority of the state-of-the-art interpretable ML methods are domain-agnostic and have evolved from fields such as computer vision, automated reasoning or statistics, making direct application to bioinformatics problems challenging without customization and domain adaptation. In this paper, we discuss the importance of explainability and algorithmic transparency in the context of bioinformatics. We provide an overview of model-specific and model-agnostic interpretable ML methods and tools and outline their potential limitations. We discuss how existing interpretable ML methods can be customized and fit to bioinformatics research problems. Further, through case studies in bioimaging, cancer genomics and text mining, we demonstrate how XAI methods can improve transparency and decision fairness. Our review aims at providing valuable insights and serving as a starting point for researchers wanting to enhance explainability and decision transparency while solving bioinformatics problems. GitHub: https://github.com/rezacsedu/XAI-for-bioinformatics.

Refining SARS-CoV-2 Intra-host Variation by Leveraging Large-scale Sequencing Data

“…Dimensionality reduction methods are valuable for distilling this data into a more manageable form (Tapinos et al 2019;Paradis 2022). However, interpreting these methods' two-dimensional representations can be challenging due to unclear biological significance (Karim et al 2022). In our workflow, we have incorporated PHATE and t-SNE alongside a metric that computes the percentage of nearest neighbours sharing the same annotation (e.g.…”

xSiGra: Explainable model for single-cell spatial data elucidation

“…However, such spatial imaging data faces challenges of missing values and data noise, which can negatively affect downstream analysis such as spatial domain detection 28, 29 . Several deep learning models have been proposed to improve noisy transcriptomics data and perform data analysis 7 , but most of them are black-box approaches that lack transparency and interpretability 18, 30 . To address this challenge, we have proposed xSiGra, which not only accurately identifies spatial cell types and enhances gene expression profiles, but also offers quantitative insights about which cells and genes are important for the identification of spatial cell types, thus making it an interpretable model.…”

“…Although the above DL-based methods prove to identify spatial cells or domains with high accuracy, their intrinsic black-box nature inhibit the explainability, regarding what genes and cells are used by these methods to achieve accurate spatial identities 17 . Such explainability issue is common when applying advanced deep learning approaches 18 . Explaining the model decisions can aid to find any limitations and validate model functioning using known knowledge 19 .…”

Toxicogenomics Approaches to Address Toxicity and Carcinogenicity in the Liver

“…Furthermore, gene expression signatures often face a “black box” problem: they frequently do not offer insight of what contributes to a positive or negative response and often do not have a biological or mechanistic link (i.e., lack of explainability). 29 This lack of transparency results in a lack of trust with the models, especially in preclinical safety organizations where explainability is often required. Finally, toxicogenomics, like most new technologies, faced a hype problem with numerous extravagant and ridiculous claims that negatively affected its ongoing adoption.…”