Introduction to AI and Machine Learning in Bioinformatics
AI and Machine Learning techniques are increasingly being used in the field of Bioinformatics to analyze and interpret large amounts of biological data. Bioinformatics is the interdisciplinary field that combines biology, computer science, and statistics to develop algorithms and software tools to analyze and interpret biological data. Machine Learning, a subset of AI, has shown great promise in analyzing complex biological data, identifying patterns, and predicting outcomes.
History of AI and Machine Learning in Bioinformatics
The use of AI and Machine Learning in Bioinformatics dates back to the early 1990s when researchers started developing algorithms to analyze DNA sequences. Since then, the field has grown significantly, and AI and Machine Learning techniques are now being used to analyze a wide range of biological data, including gene expression, protein structure, and function, and clinical data.
XAI for Bioinformatics
Explainable AI (XAI) is an emerging field that focuses on making AI and Machine Learning algorithms more transparent and interpretable. In the domain of Bioinformatics, XAI is becoming increasingly important, as it allows researchers to understand how AI algorithms make predictions and identify patterns in biological data. XAI techniques such as feature importance analysis and model interpretability are being used to analyze complex biological data and improve the accuracy and reliability of AI-based predictions.
XAI techniques are also being used to analyze and interpret clinical data, enabling healthcare professionals to make better decisions based on AI-based predictions. For example, XAI can be used to analyze patient data and identify risk factors for diseases such as cancer and heart disease. By understanding how AI algorithms make predictions, healthcare professionals can develop more effective treatment plans and improve patient outcomes.
In summary, the use of AI and Machine Learning in Bioinformatics has revolutionized the way we analyze and interpret biological data. With the increasing adoption of XAI techniques, we can expect to see more accurate, reliable, and interpretable AI-based predictions in the field of Bioinformatics, with significant benefits for healthcare and the life sciences.
In this article we find that with the development in the field of neural networks, explainable AI (XAI), is being studied to ensure that artificial intelligence models can be explained. There are some attempts to apply neural networks to neuroscientific studies to explain neurophysiological information with high machine learning performances. However, most of those studies have simply visualized features extracted from XAI and seem to lack an active neuroscientific interpretation of those features. In this study, we have tried to actively explain the high-dimensional learning features contained in the neurophysiological information extracted from XAI, compared with the previously reported neuroscientific results. Approach. We designed a deep neural network classifier using 3D information (3D DNN) and a 3D class activation map (3D CAM) to visualize high-dimensional classification features. We used those tools to classify monkey electrocorticogram (ECoG) data obtained from the unimanual and bimanual movement experiment. Main results. The 3D DNN showed better classification accuracy than other machine learning techniques, such as 2D DNN. Unexpectedly, the activation weight in the 3D CAM analysis was high in the ipsilateral motor and somatosensory cortex regions, whereas the gamma-band power was activated in the contralateral areas during unimanual movement, which suggests that the brain signal acquired from the motor cortex contains information about both contralateral movement and ipsilateral movement. Moreover, the hand-movement classification system used critical temporal information at movement onset and offset when classifying bimanual movements. Significance. As far as we know, this is the first study to use high-dimensional neurophysiological information (spatial, spectral, and temporal) with the deep learning method, reconstruct those features, and explain how the neural network works. We expect that our methods can be widely applied and used in neuroscience and electrophysiology research from the point of view of the explainability of XAI as well as its performance.
2. Applied machine learning in cancer research: A systematic review for patient diagnosis, classification and prognosis
. Artificial Intelligence (AI) has recently altered the landscape of cancer research and medical oncology using traditional Machine Learning (ML) algorithms and cutting-edge Deep Learning (DL) architectures. In this review article we focus on the ML aspect of AI applications in cancer research and present the most indicative studies with respect to the ML algorithms and data used. The PubMed and dblp databases were considered to obtain the most relevant research works of the last five years. Based on a comparison of the proposed studies and their research clinical outcomes concerning the medical ML application in cancer research, three main clinical scenarios were identified. We give an overview of the well-known DL and Reinforcement Learning (RL) methodologies, as well as their application in clinical practice, and we briefly discuss Systems Biology in cancer research. We also provide a thorough examination of the clinical scenarios with respect to disease diagnosis, patient classification and cancer prognosis and survival. The most relevant studies identified in the preceding year are presented along with their primary findings. Furthermore, we examine the effective implementation and the main points that need to be addressed in the direction of robustness, explainability and transparency of predictive models. Finally, we summarize the most recent advances in the field of AI/ML applications in cancer research and medical oncology, as well as some of the challenges and open issues that need to be addressed before data-driven models can be implemented in healthcare systems to assist physicians in their daily practice.
In Fig. 1, we present the results of our literature overview on cancer diagnosis, prognosis and patients’ risk stratification the last five years on both databases. In Fig. 1 (upper part), a group-based barplot illustrates the number of articles that were identified when considering each search query in the databases. Fig. 1 (bottom part) illustrates the timeline results for each database. The total number of articles as regards to the total sum of the articles within the three queries is depicted per year.
Fig. 1:
The current adoption of Medical Artificial Intelligence (AI) solutions in clinical practice suggest that despite its undeniable potential AI is not achieving this potential. A major barrier to its adoption is the lack of transparency and interpretability, and the inability of the system to explain its results. Explainable AI (XAI) is an emerging field in AI that aims to address these barriers, with the development of new or modified algorithms to enable transparency, provide explanations in a way that humans can understand and foster trust. Numerous XAI techniques have been proposed in the literature, commonly classified as model-agnostic or model-specific. In this study, we examine the application of four model-agnostic XAI techniques (LIME, SHAP, ANCHORS, inTrees) to an XGBoost classifier trained on real-life medical data for the prediction of high-risk asymptomatic carotid plaques based on ultrasound image analysis. We present and compare local explanations for selected observations in the test set. We also present global explanations generated from these techniques that explain the behavior of the entire model. Additionally, we assess the quality of the explanations, using suggested properties in the literature. Finally, we discuss the results of this comparative study and suggest directions for future work.
4. An Causal XAI Diagnostic Model for Breast Cancer Based on Mammography Reports
Breast cancer has become one of the most common malignant tumors in women worldwide, and it seriously threatens women’s physical and mental health. In recent years, with the development of Artificial Intelligence(AI) and the accumulation of medical data, AI has begun to be deeply integrated with mammography, MRI, ultrasound, etc. to assist physicians in disease diagnosis. However, the existing breast cancer diagnosis model based on Computer Vision(CV) is greatly affected by the image quality; on the other hand, the breast cancer diagnosis model based on Natural Language Processing(NLP) cannot effectively extract the semantic information of the mammography report. The lack of model interpretability also makes the existing diagnostic models have low confidence. In this paper, we proposed Breast Cancer Causal XAI Diagnostic Model(BCCXDM). Specifically, we first structured the mammography report. Then find the causal graph based on the structured table. We combine the existing tabular learning method TabNet with causal graphs(Causal-TabNet) to enable reasoning in the graphs to preserve the correlation between features. More importantly, we use GNN and node transition probability to aggregate node information. We evaluate our model on the real-world mammography report, and compare it with other popular interpretable methods. The experimental results show that our interpretable results are closer to the diagnostic criteria of clinicians.
The interpretability of the model can make the prediction more convincing, and it can also give the reasons for the model’s prediction errors. Therefore, while the model is required to make correct judgments of benign and malignant, it also requires the model to have a certain interpretability.
5. Detection and classification of neurons and glial cells in the MADM mouse brain using RetinaNet
The functional role of a cell is highly dependent on its gene expression, local environment, and external cues [1,2]. The two latter factors are directly related to the spatial location of the cell (e.g., layers in the neocortex, regions in the hippocampus and more). Therefore, in a tissue section, both the number of labeled cells and their spatial distribution are of paramount importance [3–5]. To build these spatial distribution maps, two technologies have been instrumental: (i) Sophisticated and automated microscopes (e.g., slide scanners) that facilitate high-throughput data acquisition of biomedical specimens. (ii) Tissue labeling methods that include immunohistochemistry, in situ hybridization, transgenic reporter mice and more.
Here our goal is to detect and classify cells in images of brain sections obtained from Mosaic Analysis with Double Markers (MADM) mice. MADM allows for simultaneous labeling and genetic manipulation in developmentally derived clones of somatic cells. We and others have extensively used MADM alleles in developmental studies on the roles of various genetic factors, which are involved with neurogenesis and gliogenesis. An advantage of MADM is that neurons and glia with distinct genotypes are permanently labeled by expression of two fluorescent proteins. Furthermore, MADM labeling occurs in sparse populations such that the entire morphology of individual cells can be easily resolved using microscopy. However, in some MADM preparations an entire brain section can contain large numbers of cells despite the sparsity of genetic labeling relative to the total number of cells, which can render manual cell counting tedious and error prone. Hence, the automation of cell detection and classification is vital to boost throughput and unbiased approaches necessary for quantification of complex tissues such as MADM brain sections.
Confocal micrograph of a sagittal section from a month-old Nestin-cre, MADM-11 (MADM) mouse forebrain (left) and the corresponding annotated map from the Allen Brain Atlas (right; Image credit: Allen Institute). Scale bar, 1500 μm. GFP–green fluorescent protein, RFP–red fluorescent protein, DAPI– 4’,6-diamidino-2-phenylindole. (B) Three isolated neurons captured in the MADM brain where green (enhanced GFP), red (tdTomato) and yellow (both reporters expressed) cells are derived from distinct clones of progenitors earlier during development. Scale bars, 40 μm. (C) A representative coronal MADM section with only a single clone of cells labeled and later imaged by a slide scanner (left; scale bar, 1000 μm). Sections were obtained from MADM brains in which red, green, and yellow clones were labeled in the late-stage embryo at very low densities using a Nestin-creER transgene. Boxed area demarcates the zoomed image on the right. Scale bar, 50 μm. (D) A representative confocal image of another MADM brain (left; scale bar, 1250 μm) and zoomed image to the right (right; scale bar, 50 μm). Two main types of cells can be seen in C and D: Neurons and glia marked with arrowheads and arrows, respectively. The white dashed frame in D indicates a glia cluster. (E) The cell detection workflow. To localize and classify each cell, an object detection network (RetinaNet) was utilized. To detect dense and saturated glia clusters, two RetinaNet models were trained, one to detect individual cells with different colors, and the other to detect only glia clusters. In the inference stage, predictions of individual cells and glial clusters were merged to obtain final output.
6. A novel interaction-based methodology towards explainable AI with better understanding of Pneumonia Chest X-ray Images
Many successful achievements in machine learning and deep learning have accelerated real-world implementations of Artificial Intelligence (AI). This issue has been greatly acknowledged by the Department of Defense (DoD). DARPA initiated the eXplainable Artificial Intelligence (XAI) challenge and brought this new interest to the surface. In addressing the concepts of interpretability and explainability, these scholars and researchers have made attempts towards discussing a trade-off between learning performance (usually measured by prediction performance) and effectiveness of explanations (also known as explainability), which is presented in . This trade-off often occurs in any supervised machine learning problems that aim to use explanatory variable to predict response variable (or outcome variable) which happens between learning performance (also known as prediction performance) and effectiveness of explanations (also known as explainability). As illustrated in, the issue is that a learning algorithm such as linear regression modeling has a clear algorithmic structure and an explicitly written mathematical expression so that it can be understood with high effectiveness of explanations with yet relatively lower prediction performance. An algorithm such as linear regression can be positioned in the bottom right corner of the scale in this figure (in consensus, linear regression is regarded as an explainable learning algorithm). On the other hand, a learning algorithm such as a deep Convolutional Neural Network (CNN) with hundreds of millions of parameters would have much better prediction performance, yet it is extremely challenging to explicitly state the mathematical formulation of the architecture. A deep learning algorithm such as an ultra deep CNN with hundreds of millions of parameters would be positioned on the top left corner of the scale in the figure (which is generally considered inexplainable in consensus). In the field of transfer learning, it is a common practice to adopt a previously trained CNN model on a new data set. For example, one can adopt the VGG16 model and weights learned from ImageNet on a new data: Chest X-ray Images. The filters learned from ImageNet data may or may not be helpful on Chest X-ray. Due to large amount of filters used in VGG16, we can hope that some filters can capture important information on Chest X-ray scans. However, we will never truly know what features are important if we do not impose any feature assessment condition. This renders the adoption of a pretrained CNN model inexplicable. This calls for the need of a novel feature assessment and feature selection technique to shrink the dimension of the number of parameters while maintaining prediction performance. Hence, this paper focuses on feature and variable selection assessment to build explainability including trustworthy, fair, robust, and high performing models for real-world applications.
A popular description of interpretability defines XAI as the ability to explain or to present in understandable terms to a human Doshi-Velez and Kim. Another popular version states interpretability as the degree to which a human can understand the cause of a decision. Though intuitive, these definitions lack mathematical formality and rigorousness. Moreover, it is yet unclear why variables provide us the good prediction performance and, more importantly, how to yield a relatively unbiased estimate of a parameter that is not sensitive to noisy variables and is related to the parameter of interest.
To shed light to these questions, we define the following three necessary conditions (, , and ) for any feature selection methodology to be explainable and interpretable. In other words, a variable and feature selection method can only be considered explainable and interpretable if all three conditions (, , and defined below) are satisfied. Specifically, we regard the final assessment quantity of the importance evaluated for a set of features or variables to be the final score measured for feature assessment and selection method. More importantly, we define this importance score of a variable set from using only explainable feature assessment and selection methods to be the explainability of a combination of variables. There are three conditions defined below and we name these conditions , , and .
- . The first condition states that the feature selection methodology must be non-parametric and hence does not require any assumption of the true form of the real model.
- . An explainable and interpretable feature selection method must clearly state to what degree a combination of explanatory variables influences the response variable.
- . In order for a feature assessment and selection technique to be interpretable and explainable, it must related with the predictivity of the explanatory variables.
The first few months of 2020 have profoundly changed the way we live our lives and carry out our daily activities. Although the widespread use of futuristic robotaxis and self-driving commercial vehicles has not yet become a reality, the COVID-19 pandemic has dramatically accelerated the adoption of Artificial Intelligence (AI) in different fields. We have witnessed the equivalent of two years of digital transformation compressed into just a few months. Whether it is in tracing epidemiological peaks or in transacting contactless payments, the impact of these developments has been almost immediate, and a window has opened up on what is to come. Here we analyze and discuss how AI can support us in facing the ongoing pandemic. Despite the numerous and undeniable contributions of AI, clinical trials and human skills are still required. Even if different strategies have been developed in different states worldwide, the fight against the pandemic seems to have found everywhere a valuable ally in AI, a global and open-source tool capable of providing assistance in this health emergency. A careful AI application would enable us to operate within this complex scenario involving healthcare, society and research.
COVID-19 may be considered the first influenza pandemic to be disseminated in our hyper-connected world. It has proven to be a phenomenon that significantly and rapidly impacts many layers of our society. Despite the many containment measures adopted to limit COVID transmissions, such as the closing of borders and the introduction of periods of lockdown, we are witnessing as many as 116 million confirmed cases and more than 2 million deaths in 235 different countries, as reported by the World Organization Health (WHO) at the end of February 2021. Serious concerns about healthcare systems’ capacity have arisen due to the unprecedented demand for health services, especially concerning disadvantaged states. In this scenario, methodologies able to speed up diagnostic procedures, enhance monitoring and tracking capabilities, predict the evolutionary stages of the contagion as well as its effects on society, and simulate the results of a containment strategy, a medical protocol or a new molecule, can represent a revolutionary milestone in the progress of the world in facing these dramatic events. The COVID-19 emergency has given an incredible boost to the improvement of existing models and the development of new prototypes in order to achieve promising results in fields such as the tracing of the infection (Pinotti et al. 2020) or the prediction of its diffusion and the effects of restrictive measures (Della Rossa et al. 2020). Advances in AI are expected to represent an effective strategy to face these challenges: thanks to the massive amount of information made available through the advent of pervasive IT, and the continually increasing computational power, AI has shown an outstanding performance concerning most of the problems mentioned above. Indeed, its ability to extract patterns and relations from data has made this research area particularly attractive in tasks involving the description of complex information and dynamics.
8. Explainable AI for Bioinformatics: Methods, Tools, and Applications
Handling large-scale biomedical data involves significant challenges, including heterogeneity, high dimensionality, unstructured data, and high levels of noise and uncertainty. Despite its data-driven nature and technical complexity, the adoption of data-driven approaches in many bioinformatics scenarios is hindered by the lack of efficient ML models capable of tackling these challenges. In order for AI systems to provide trustworthy and reliable decisions, the need for interpretable ML models has become increasingly important to ensure transparency, fairness, and accountability in critical situations. Although not all predictions need to be explained, having a model that is interpretable can make it easier for users to understand and trust its decisions. Weber et al. showed via some experiments that under the right conditions, augmentations based on XAI can provide significant, diverse, and reliable benefits advantages over black-box models. We outline some key benefits of interpretable ML methods in fig. 1. Helps avoid practical consequences One of the critical applications of AI is aiding diagnosis and treatment of various cancerous conditions. Early detection and classification of patients into high or low-risk groups is crucial for effective management of the illness. An example of this is a doctor diagnosing a patient with breast cancer.
Given that breast cancer is a leading cause of death in women, it is important for the diagnosis to be thoroughly investigated. By utilizing omics data, such as genetic mutations, copy number variations (CNVs), gene expression (GE), DNA methylation, and miRNA expression, accurate diagnosis and treatment can be identified. Suppose a deep learning model trained on multiomics data can classify cancerous samples from healthy samples with 95% accuracy, and the model diagnoses a patient with a 70% probability of having breast cancer. If the patient asks questions like “why do I have breast cancer?" or “how did the model reach this decision?" or “which biomarkers are responsible?", it may not be possible to clearly explain the model’s decision-making process because the representations learned by the deep learning model may not be easily interpretable. The diagnosis may further depend on several distinct molecular subtypes and factors like estrogen-, progesterone-, and human epidermal growth factor receptors. The diagnosis of a breast cancer patient requires a careful examination of multiple sources of data, including omics information, bioimaging, and clinical records. A multimodal DNN model trained on this data can classify samples with high accuracy. Further, image-guided pathology may need to employ in order to analyze imaging data (e.g., histopathological image analysis), as shown in fig. 2. However, the representations and decision-making process of such a multimodal model may not be easily interpreted.
Bibliography
[0] Kingsmore, Kathryn M; Puglisi, Christopher E; Grammer, Amrie C; Lipsky, Peter E. Nature Reviews. Rheumatology; London Vol. 17, Iss. 12, (Dec 2021): 710-730. DOI:10.1038/s41584-021-00708-w
An introduction to machine learning and analysis of its use in rheumatic diseases
Published 25 November 2021 • © 2021 IOP Publishing Ltd
