Explainable AI for Bioinformatics: Methods, Tools and Applications.

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.

Deep learning-based clustering approaches for bioinformatics.

Clustering is central to many data-driven bioinformatics research and serves a powerful computational method. In particular, clustering helps at analyzing unstructured and high-dimensional data in the form of sequences, expressions, texts and images. Further, clustering is used to gain insights into biological processes in the genomics level, e.g. clustering of gene expressions provides insights on the natural structure inherent in the data, understanding gene functions, cellular processes, subtypes of cells and understanding gene regulations. Subsequently, clustering approaches, including hierarchical, centroid-based, distribution-based, density-based and self-organizing maps, have long been studied and used in classical machine learning settings. In contrast, deep learning (DL)-based representation and feature learning for clustering have not been reviewed and employed extensively. Since the quality of clustering is not only dependent on the distribution of data points but also on the learned representation, deep neural networks can be effective means to transform mappings from a high-dimensional data space into a lower-dimensional feature space, leading to improved clustering results. In this paper, we review state-of-the-art DL-based approaches for cluster analysis that are based on representation learning, which we hope to be useful, particularly for bioinformatics research. Further, we explore in detail the training procedures of DL-based clustering algorithms, point out different clustering quality metrics and evaluate several DL-based approaches on three bioinformatics use cases, including bioimaging, cancer genomics and biomedical text mining. We believe this review and the evaluation results will provide valuable insights and serve a starting point for researchers wanting to apply DL-based unsupervised methods to solve emerging bioinformatics research problems.

ReVeaLD: a user-driven domain-specific interactive search platform for biomedical research.

Bioinformatics research relies heavily on the ability to discover and correlate data from various sources. The specialization of life sciences over the past decade, coupled with an increasing number of biomedical datasets available through standardized interfaces, has created opportunities towards new methods in biomedical discovery. Despite the popularity of semantic web technologies in tackling the integrative bioinformatics challenge, there are many obstacles towards its usage by non-technical research audiences. In particular, the ability to fully exploit integrated information needs using improved interactive methods intuitive to the biomedical experts. In this report we present ReVeaLD (a Real-time Visual Explorer and Aggregator of Linked Data), a user-centered visual analytics platform devised to increase intuitive interaction with data from distributed sources. ReVeaLD facilitates query formulation using a domain-specific language (DSL) identified by biomedical experts and mapped to a self-updated catalogue of elements from external sources. ReVeaLD was implemented in a cancer research setting; queries included retrieving data from in silico experiments, protein modeling and gene expression. ReVeaLD was developed using Scalable Vector Graphics and JavaScript and a demo with explanatory video is available at http://www.srvgal78.deri.ie:8080/explorer. A set of user-defined graphic rules controls the display of information through media-rich user interfaces. Evaluation of ReVeaLD was carried out as a game: biomedical researchers were asked to assemble a set of 5 challenge questions and time and interactions with the platform were recorded. Preliminary results indicate that complex queries could be formulated under less than two minutes by unskilled researchers. The results also indicate that supporting the identification of the elements of a DSL significantly increased intuitiveness of the platform and usability of semantic web technologies by domain users.

Facilitating the Sharing of Electrophysiology Data Analysis Results Through In-Depth Provenance Capture.

Scientific research demands reproducibility and transparency, particularly in data-intensive fields like electrophysiology. Electrophysiology data are typically analyzed using scripts that generate output files, including figures. Handling these results poses several challenges due to the complexity and iterative nature of the analysis process. These stem from the difficulty to discern the analysis steps, parameters, and data flow from the results, making knowledge transfer and findability challenging in collaborative settings. Provenance information tracks data lineage and processes applied to it, and provenance capture during the execution of an analysis script can address those challenges. We present Alpaca (Automated Lightweight Provenance Capture), a tool that captures fine-grained provenance information with minimal user intervention when running data analysis pipelines implemented in Python scripts. Alpaca records inputs, outputs, and function parameters and structures information according to the W3C PROV standard. We demonstrate the tool using a realistic use case involving multichannel local field potential recordings of a neurophysiological experiment, highlighting how the tool makes result details known in a standardized manner in order to address the challenges of the analysis process. Ultimately, using Alpaca will help to represent results according to the FAIR principles, which will improve research reproducibility and facilitate sharing the results of data analyses.

Improving the Quality of Unstructured Cancer Data Using Large Language Models: A German Oncological Case Study.

With cancer being a leading cause of death globally, epidemiological and clinical cancer registration is paramount for enhancing oncological care and facilitating scientific research. However, the heterogeneous landscape of medical data presents significant challenges to the current manual process of tumor documentation. This paper explores the potential of Large Language Models (LLMs) for transforming unstructured medical reports into the structured format mandated by the German Basic Oncology Dataset. Our findings indicate that integrating LLMs into existing hospital data management systems or cancer registries can significantly enhance the quality and completeness of cancer data collection - a vital component for diagnosing and treating cancer and improving the effectiveness and benefits of therapies. This work contributes to the broader discussion on the potential of artificial intelligence or LLMs to revolutionize medical data processing and reporting in general and cancer care in particular.

A study on interoperability between two Personal Health Train infrastructures in leukodystrophy data analysis.

The development of platforms for distributed analytics has been driven by a growing need to comply with various governance-related or legal constraints. Among these platforms, the so-called Personal Health Train (PHT) is one representative that has emerged over the recent years. However, in projects that require data from sites featuring different PHT infrastructures, institutions are facing challenges emerging from the combination of multiple PHT ecosystems, including data governance, regulatory compliance, or the modification of existing workflows. In these scenarios, the interoperability of the platforms is preferable. In this work, we introduce a conceptual framework for the technical interoperability of the PHT covering five essential requirements: Data integration, unified station identifiers, mutual metadata, aligned security protocols, and business logic. We evaluated our concept in a feasibility study that involves two distinct PHT infrastructures: PHT-meDIC and PADME. We analyzed data on leukodystrophy from patients in the University Hospitals of Tübingen and Leipzig, and patients with differential diagnoses at the University Hospital Aachen. The results of our study demonstrate the technical interoperability between these two PHT infrastructures, allowing researchers to perform analyses across the participating institutions. Our method is more space-efficient compared to the multi-homing strategy, and it shows only a minimal time overhead.

Will it run?-A proof of concept for smoke testing decentralized data analytics experiments.

The growing interest in data-driven medicine, in conjunction with the formation of initiatives such as the European Health Data Space (EHDS) has demonstrated the need for methodologies that are capable of facilitating privacy-preserving data analysis. Distributed Analytics (DA) as an enabler for privacy-preserving analysis across multiple data sources has shown its potential to support data-intensive research. However, the application of DA creates new challenges stemming from its distributed nature, such as identifying single points of failure (SPOFs) in DA tasks before their actual execution. Failing to detect such SPOFs can, for example, result in improper termination of the DA code, necessitating additional efforts from multiple stakeholders to resolve the malfunctions. Moreover, these malfunctions disrupt the seamless conduct of DA and entail several crucial consequences, including technical obstacles to resolve the issues, potential delays in research outcomes, and increased costs. In this study, we address this challenge by introducing a concept based on a method called Smoke Testing, an initial and foundational test run to ensure the operability of the analysis code. We review existing DA platforms and systematically extract six specific Smoke Testing criteria for DA applications. With these criteria in mind, we create an interactive environment called Development Environment for AuTomated and Holistic Smoke Testing of Analysis-Runs (DEATHSTAR), which allows researchers to perform Smoke Tests on their DA experiments. We conduct a user-study with 29 participants to assess our environment and additionally apply it to three real use cases. The results of our evaluation validate its effectiveness, revealing that 96.6% of the analyses created and (Smoke) tested by participants using our approach successfully terminated without any errors. Thus, by incorporating Smoke Testing as a fundamental method, our approach helps identify potential malfunctions early in the development process, ensuring smoother data-driven research within the scope of DA. Through its flexibility and adaptability to diverse real use cases, our solution enables more robust and efficient development of DA experiments, which contributes to their reliability.

Convolutional Embedded Networks for Population Scale Clustering and Bio-Ancestry Inferencing.

The study of genetic variants (GVs) can help find correlating population groups and to identify cohorts that are predisposed to common diseases and explain differences in disease susceptibility and how patients react to drugs. Machine learning techniques are increasingly being applied to identify interacting GVs to understand their complex phenotypic traits. Since the performance of a learning algorithm not only depends on the size and nature of the data but also on the quality of underlying representation, deep neural networks (DNNs) can learn non-linear mappings that allow transforming GVs data into more clustering and classification friendly representations than manual feature selection. In this paper, we propose convolutional embedded networks (CEN) in which we combine two DNN architectures called convolutional embedded clustering (CEC) and convolutional autoencoder (CAE) classifier for clustering individuals and predicting geographic ethnicity based on GVs, respectively. We employed CAE-based representation learning to 95 million GVs from the '1000 genomes' (covering 2,504 individuals from 26 ethnic origins) and 'Simons genome diversity' (covering 279 individuals from 130 ethnic origins) projects. Quantitative and qualitative analyses with a focus on accuracy and scalability show that our approach outperforms state-of-the-art approaches such as VariantSpark and ADMIXTURE. In particular, CEC can cluster targeted population groups in 22 hours with an adjusted rand index (ARI) of 0.915, the normalized mutual information (NMI) of 0.92, and the clustering accuracy (ACC) of 89 percent. Contrarily, the CAE classifier can predict the geographic ethnicity of unknown samples with an F1 and Mathews correlation coefficient (MCC) score of 0.9004 and 0.8245, respectively. Further, to provide interpretations of the predictions, we identify significant biomarkers using gradient boosted trees (GBT) and SHapley Additive exPlanations (SHAP). Overall, our approach is transparent and faster than the baseline methods, and scalable for 5 to 100 percent of the full human genome.

A Privacy-Preserving Distributed Analytics Platform for Health Care Data.

BACKGROUND: In recent years, data-driven medicine has gained increasing importance in terms of diagnosis, treatment, and research due to the exponential growth of health care data. However, data protection regulations prohibit data centralisation for analysis purposes because of potential privacy risks like the accidental disclosure of data to third parties. Therefore, alternative data usage policies, which comply with present privacy guidelines, are of particular interest. OBJECTIVE: We aim to enable analyses on sensitive patient data by simultaneously complying with local data protection regulations using an approach called the Personal Health Train (PHT), which is a paradigm that utilises distributed analytics (DA) methods. The main principle of the PHT is that the analytical task is brought to the data provider and the data instances remain in their original location. METHODS: In this work, we present our implementation of the PHT paradigm, which preserves the sovereignty and autonomy of the data providers and operates with a limited number of communication channels. We further conduct a DA use case on data stored in three different and distributed data providers. RESULTS: We show that our infrastructure enables the training of data models based on distributed data sources. CONCLUSION: Our work presents the capabilities of DA infrastructures in the health care sector, which lower the regulatory obstacles of sharing patient data. We further demonstrate its ability to fuel medical science by making distributed data sets available for scientists or health care practitioners.

Distributed Skin Lesion Analysis Across Decentralised Data Sources.

Skin cancer has become the most common cancer type. Research has applied image processing and analysis tools to support and improve the diagnose process. Conventional procedures usually centralise data from various data sources to a single location and execute the analysis tasks on central servers. However, centralisation of medical data does not often comply with local data protection regulations due to its sensitive nature and the loss of sovereignty if data providers allow unlimited access to the data. The Personal Health Train (PHT) is a Distributed Analytics (DA) infrastructure bringing the algorithms to the data instead of vice versa. By following this paradigm shift, it proposes a solution for persistent privacy- related challenges. In this work, we present a feasibility study, which demonstrates the capability of the PHT to perform statistical analyses and Machine Learning on skin lesion data distributed among three Germany-wide data providers.

Enabling ad-hoc reuse of private data repositories through schema extraction.

BACKGROUND: Sharing sensitive data across organizational boundaries is often significantly limited by legal and ethical restrictions. Regulations such as the EU General Data Protection Rules (GDPR) impose strict requirements concerning the protection of personal and privacy sensitive data. Therefore new approaches, such as the Personal Health Train initiative, are emerging to utilize data right in their original repositories, circumventing the need to transfer data. RESULTS: Circumventing limitations of previous systems, this paper proposes a configurable and automated schema extraction and publishing approach, which enables ad-hoc SPARQL query formulation against RDF triple stores without requiring direct access to the private data. The approach is compatible with existing Semantic Web-based technologies and allows for the subsequent execution of such queries in a safe setting under the data provider's control. Evaluation with four distinct datasets shows that a configurable amount of concise and task-relevant schema, closely describing the structure of the underlying data, was derived, enabling the schema introspection-assisted authoring of SPARQL queries. CONCLUSIONS: Automatically extracting and publishing data schema can enable the introspection-assisted creation of data selection and integration queries. In conjunction with the presented system architecture, this approach can enable reuse of data from private repositories and in settings where agreeing upon a shared schema and encoding a priori is infeasible. As such, it could provide an important step towards reuse of data from previously inaccessible sources and thus towards the proliferation of data-driven methods in the biomedical domain.

Design of Metadata Services for Clinical Data Interoperability in Germany.

Secondary use of electronic health record (EHR) data requires a detailed description of metadata, especially when data collection and data re-use are organizationally and technically far apart. This paper describes the concept of the SMITH consortium that includes conventions, processes, and tools for describing and managing metadata using common standards for semantic interoperability. It deals in particular with the chain of processing steps of data from existing information systems and provides an overview of the planned use of metadata, medical terminologies, and semantic services in the consortium.

Smart Medical Information Technology for Healthcare (SMITH).

INTRODUCTION: This article is part of the Focus Theme of Methods of Information in Medicine on the German Medical Informatics Initiative. "Smart Medical Information Technology for Healthcare (SMITH)" is one of four consortia funded by the German Medical Informatics Initiative (MI-I) to create an alliance of universities, university hospitals, research institutions and IT companies. SMITH's goals are to establish Data Integration Centers (DICs) at each SMITH partner hospital and to implement use cases which demonstrate the usefulness of the approach. OBJECTIVES: To give insight into architectural design issues underlying SMITH data integration and to introduce the use cases to be implemented. GOVERNANCE AND POLICIES: SMITH implements a federated approach as well for its governance structure as for its information system architecture. SMITH has designed a generic concept for its data integration centers. They share identical services and functionalities to take best advantage of the interoperability architectures and of the data use and access process planned. The DICs provide access to the local hospitals' Electronic Medical Records (EMR). This is based on data trustee and privacy management services. DIC staff will curate and amend EMR data in the Health Data Storage. METHODOLOGY AND ARCHITECTURAL FRAMEWORK: To share medical and research data, SMITH's information system is based on communication and storage standards. We use the Reference Model of the Open Archival Information System and will consistently implement profiles of Integrating the Health Care Enterprise (IHE) and Health Level Seven (HL7) standards. Standard terminologies will be applied. The SMITH Market Place will be used for devising agreements on data access and distribution. 3LGM2 for enterprise architecture modeling supports a consistent development process.The DIC reference architecture determines the services, applications and the standardsbased communication links needed for efficiently supporting the ingesting, data nourishing, trustee, privacy management and data transfer tasks of the SMITH DICs. The reference architecture is adopted at the local sites. Data sharing services and the market place enable interoperability. USE CASES: The methodological use case "Phenotype Pipeline" (PheP) constructs algorithms for annotations and analyses of patient-related phenotypes according to classification rules or statistical models based on structured data. Unstructured textual data will be subject to natural language processing to permit integration into the phenotyping algorithms. The clinical use case "Algorithmic Surveillance of ICU Patients" (ASIC) focusses on patients in Intensive Care Units (ICU) with the acute respiratory distress syndrome (ARDS). A model-based decision-support system will give advice for mechanical ventilation. The clinical use case HELP develops a "hospital-wide electronic medical record-based computerized decision support system to improve outcomes of patients with blood-stream infections" (HELP). ASIC and HELP use the PheP. The clinical benefit of the use cases ASIC and HELP will be demonstrated in a change of care clinical trial based on a step wedge design. DISCUSSION: SMITH's strength is the modular, reusable IT architecture based on interoperability standards, the integration of the hospitals' information management departments and the public-private partnership. The project aims at sustainability beyond the first 4-year funding period.

BioFed: federated query processing over life sciences linked open data.

BACKGROUND: Biomedical data, e.g. from knowledge bases and ontologies, is increasingly made available following open linked data principles, at best as RDF triple data. This is a necessary step towards unified access to biological data sets, but this still requires solutions to query multiple endpoints for their heterogeneous data to eventually retrieve all the meaningful information. Suggested solutions are based on query federation approaches, which require the submission of SPARQL queries to endpoints. Due to the size and complexity of available data, these solutions have to be optimised for efficient retrieval times and for users in life sciences research. Last but not least, over time, the reliability of data resources in terms of access and quality have to be monitored. Our solution (BioFed) federates data over 130 SPARQL endpoints in life sciences and tailors query submission according to the provenance information. BioFed has been evaluated against the state of the art solution FedX and forms an important benchmark for the life science domain. METHODS: The efficient cataloguing approach of the federated query processing system 'BioFed', the triple pattern wise source selection and the semantic source normalisation forms the core to our solution. It gathers and integrates data from newly identified public endpoints for federated access. Basic provenance information is linked to the retrieved data. Last but not least, BioFed makes use of the latest SPARQL standard (i.e., 1.1) to leverage the full benefits for query federation. The evaluation is based on 10 simple and 10 complex queries, which address data in 10 major and very popular data sources (e.g., Dugbank, Sider). RESULTS: BioFed is a solution for a single-point-of-access for a large number of SPARQL endpoints providing life science data. It facilitates efficient query generation for data access and provides basic provenance information in combination with the retrieved data. BioFed fully supports SPARQL 1.1 and gives access to the endpoint's availability based on the EndpointData graph. Our evaluation of BioFed against FedX is based on 20 heterogeneous federated SPARQL queries and shows competitive execution performance in comparison to FedX, which can be attributed to the provision of provenance information for the source selection. CONCLUSION: Developing and testing federated query engines for life sciences data is still a challenging task. According to our findings, it is advantageous to optimise the source selection. The cataloguing of SPARQL endpoints, including type and property indexing, leads to efficient querying of data resources over the Web of Data. This could even be further improved through the use of ontologies, e.g., for abstract normalisation of query terms.

TopFed: TCGA tailored federated query processing and linking to LOD.

BACKGROUD: The Cancer Genome Atlas (TCGA) is a multidisciplinary, multi-institutional effort to catalogue genetic mutations responsible for cancer using genome analysis techniques. One of the aims of this project is to create a comprehensive and open repository of cancer related molecular analysis, to be exploited by bioinformaticians towards advancing cancer knowledge. However, devising bioinformatics applications to analyse such large dataset is still challenging, as it often requires downloading large archives and parsing the relevant text files. Therefore, it is making it difficult to enable virtual data integration in order to collect the critical co-variates necessary for analysis. METHODS: We address these issues by transforming the TCGA data into the Semantic Web standard Resource Description Format (RDF), link it to relevant datasets in the Linked Open Data (LOD) cloud and further propose an efficient data distribution strategy to host the resulting 20.4 billion triples data via several SPARQL endpoints. Having the TCGA data distributed across multiple SPARQL endpoints, we enable biomedical scientists to query and retrieve information from these SPARQL endpoints by proposing a TCGA tailored federated SPARQL query processing engine named TopFed. RESULTS: We compare TopFed with a well established federation engine FedX in terms of source selection and query execution time by using 10 different federated SPARQL queries with varying requirements. Our evaluation results show that TopFed selects on average less than half of the sources (with 100% recall) with query execution time equal to one third to that of FedX. CONCLUSION: With TopFed, we aim to offer biomedical scientists a single-point-of-access through which distributed TCGA data can be accessed in unison. We believe the proposed system can greatly help researchers in the biomedical domain to carry out their research effectively with TCGA as the amount and diversity of data exceeds the ability of local resources to handle its retrieval and parsing.