Data Science and AI in Life Sciences: Applications and Emerging Trends

Data science and artificial intelligence now form the analytical backbone of modern life sciences research, from genomic sequencing and drug discovery to clinical decision support and crop improvement. For students, researchers, analysts, and life-science professionals, understanding where these tools add genuine value, where they fail, and how to evaluate them critically is essential for productive research workflows. This article maps the current applications of data science and AI across the life sciences, examines emerging trends, and provides a practical framework for assessing AI tools before integrating them into research pipelines.

The Current State of Data Science and AI in Life Sciences

The convergence of high-throughput biology and machine learning has produced a research environment where computational methods are no longer optional add-ons but core infrastructure. Public data repositories maintained by the National Center for Biotechnology Information provide the raw material for much of this work, while training resources from the European Bioinformatics Institute help researchers build the necessary computational skills. The scale of biological data generation has outpaced traditional analysis methods, creating demand for approaches that can identify patterns across millions of data points.

Machine learning applications in genetics and genomics have shown particular promise because these fields generate large, complex datasets that can reveal insights into disease risk, genetic disorder pathogenesis, and health prediction. However, researchers must exercise caution against biases and inflated results that can produce harmful unintended impacts. Understanding the metrics used to evaluate machine learning models is therefore not a technical footnote but a core competency that influences the critical interpretation of results. Common pitfalls during model evaluation can lead researchers to overstate the reliability of their findings, particularly when working with high-dimensional genomic data where the number of features far exceeds the number of samples.

The practical reality is that AI tools in life sciences operate within a broader ecosystem of data generation, quality control, and biological validation. A machine learning prediction is only as trustworthy as the data that feeds it, and the biological interpretation that follows. Researchers who treat AI as a black box risk producing results that are statistically impressive but biologically meaningless or clinically irrelevant.

Core Applications Across the Life Sciences

Genomics and Single-Cell Analysis

Genomic data analysis has been transformed by machine learning approaches that can handle the scale and complexity of modern sequencing data. Single-cell omics technologies now allow researchers to examine transcription profiles of individual cells, providing unprecedented resolution into cellular heterogeneity. When combined with large-scale perturbation screens that target specific biological mechanisms, these technologies enable measurement of perturbation effects on the whole transcriptome. This creates opportunities to understand the causative role of genes in complex biological processes such as gene regulation, disease progression, and cellular development.

The high-dimensional nature of single-cell data, coupled with the intricate complexity of biological systems, makes this analysis nontrivial. Causal machine learning has emerged as a promising direction for single-cell genomics, adapting established causal techniques to handle high-dimensional data. The model underlying most current causal approaches to single-cell biology rests on assumptions that must be examined from a biological point of view. Open problems include generalizing to unseen environments, learning interpretable models, and learning causal models of dynamics. Researchers are exploring both computational approaches and adaptations of experimental protocols to address these challenges. With the advent of single-cell atlases and increasing perturbation data, causal models are expected to become a crucial tool for informed experimental design.

Drug Discovery and Target Identification

Drug discovery has become one of the most active areas for AI application in life sciences. The concept of the receptorome, which describes the receptors, ion channels, and transporters in the human genome that are potential drug targets, has shaped how researchers approach target identification. These proteins comprise a considerable fraction of the human genome, including the G protein-coupled receptors that are targets for many medications. Recent advances in the field have challenged the assumption that the ultimate goal of drug discovery is the development of highly selective single-target drugs. Potential side effects can also become the goal of multi-target drug screening, and computational screening combined with public domain databases has expanded the toolkit available to investigators.

Membrane proteins play crucial physiological roles and represent the major category of drug targets for pharmaceuticals. Most drugs achieve therapeutic effects by interacting with membrane proteins, making them vital hubs in the biological network. Typical membrane protein targets include G protein-coupled receptors, transporters, and ion channels. Network servers and databases that contain drug, drug-target information, and relevant data support this work. State-of-the-art computational models for predicting drug-target interaction include network-based approaches and machine-learning-based approaches. Current achievements in this area point toward drug repurposing and drug discovery as prospective directions, with improved frameworks needed in bioactivity data, predicted approaches, and alternative understanding of drug bioactivity and biological processes.

Proton-sensing G protein-coupled receptors illustrate the specificity possible in target identification. Cells in tumor microenvironments use several mechanisms to sense low pH, including through proton-sensing G protein-coupled receptors such as GPR4, GPR65, GPR68, and GPR132. Numerous cancers show increased expression of these receptors, which may contribute to features of the malignant phenotype through actions on specific cell types in the tumor microenvironment, thereby promoting tumor survival and growth. Bioinformatics approaches can infer receptor expression in cell types within the tumor microenvironment, providing tools to define their contributions to tumor biology and identify potentially novel therapeutic agents.

Structural Biology and Protein Design

Deep learning has fundamentally changed protein structure prediction and design. Cyclic peptides have gained significant traction as a therapeutic modality, but the development of deep learning methods for accurately designing such peptides has been slow, mostly due to the lack of sufficiently large training sets. AfCycDesign, a deep learning approach for accurate structure prediction, sequence redesign, and de novo hallucination of cyclic peptides, demonstrates what is now possible. Using this approach, researchers identified over 10,000 structurally diverse designs predicted to fold into the designed structures with high confidence. X-ray crystal structures for eight tested de novo designed sequences matched the design models closely, highlighting atomic level accuracy. The set of hallucinated peptides served as starting scaffolds to design binders with nanomolar IC50 against MDM2 and Keap1, providing a basis for custom design of peptides for diverse protein targets and therapeutic applications.

Cryo-electron microscopy is transitioning from determining structures of isolated proteins in vitro to visualizing macromolecular architecture directly in situ. Conventional in situ approaches, primarily relying on cryo-electron tomography combined with subtomogram averaging, are often limited in resolution due to complex workflows, cumulative errors in processing, and low data throughput. Emerging in situ single-particle cryo-EM methods address these limitations by collecting high-dose, untilted images of cellular lamellae. Using high-resolution templates for particle identification and refinement, these methods have significantly advanced both data throughput and achievable resolution.

Clinical Data Analysis and Decision Support

Clinical decision support systems leverage health data analytics to support clinical decision-making, but systematic evidence regarding their deployment in pediatric oncology remains limited. This gap in the literature gives rise to unique challenges, particularly in addressing the individualized nature of pediatric cancer diagnosis, treatment, and the doctor-nurse-patient tripartite decision-making. A scoping review of seventeen studies revealed that clinical decision support systems demonstrate potential for application in pediatric cancer management, but the majority of applications are concentrated in the treatment phase, with comparatively less emphasis on risk assessment, diagnosis, nursing workflow, and survivorship care. User interaction and follow-up functionalities remain suboptimal and warrant further optimization. There is a critical need for multicenter randomized controlled trials to validate the clinical efficacy of these systems.

Machine learning frameworks for lung cancer prognosis and detection combine deep learning, genomics, and radiomics to enhance detection, classification, and prognosis. Advanced techniques such as support vector machines, random forest, convolutional neural networks, artificial neural networks, and XGBoost have demonstrated high accuracy in analyzing clinical and imaging data, including CT and PET scans. These models assist in identifying tumors, predicting cancer stages, and evaluating treatment responses. However, limitations such as poor generalization across populations, limited interpretability, and data imbalance still hinder broader application. The emphasis on integrating reliable and transparent AI solutions into clinical settings supports precision medicine, with future research focused on developing interpretable hybrid models, incorporating multi-modal data, and establishing regulatory standards.

Agricultural and Crop Science Applications

Machine learning has emerged as a promising tool for crop improvement, particularly for orphan crops that are important sources of nutrition in developing regions. Many orphan crops are tolerant to biotic and abiotic stressors, but modern crop improvement technologies have not been widely applied due to the lack of resources available. Transferring knowledge from major crops to orphan crops and using machine learning to improve accuracy and efficiency can advance orphan crop breeding. The conservation of genes between related species supports this knowledge transfer approach.

Machine learning applications in gas adsorption research using nanoporous materials illustrate the broader pattern of AI adoption in materials and agricultural sciences. Publication analysis shows that annual output remained generally below 20 before 2019, then increased rapidly to reach approximately 280 publications in 2025, indicating accelerated integration of machine learning with adsorption simulation, material screening, and performance evaluation. The source distribution broadened from a limited set of chemistry and engineering journals to diverse venues. Collaboration analysis identified compact author clusters, but weak bridging links among clusters indicate that cross-community collaboration remains limited.

At a Glance: AI Tool Applications and Evaluation Criteria

A Framework for Evaluating AI Tools in Research Workflows

Step 1: Define the Biological Question and Required Output

Before selecting any AI tool, specify the biological question precisely and determine what output format will be actionable. A tool that predicts protein structures serves a different purpose than one that identifies biomarkers from gene expression data. The output must connect to a downstream validation step, whether that is experimental confirmation, clinical application, or breeding decision. Researchers should document the expected output type, the tolerance for error, and the validation pathway before investing time in tool implementation.

Step 2: Assess Data Readiness and Quality

AI tools are data-hungry, and the quality of input data determines the ceiling for model performance. Assess whether your dataset has sufficient sample size, appropriate controls, and documented provenance. For genomic data, verify that sequencing depth meets the requirements of the intended analysis. For clinical data, confirm that the population represented in the training data matches the population where the tool will be applied. The Genomic Data Sharing Policy from the National Institutes of Health provides guidance on responsible data sharing practices that researchers should incorporate into their workflows.

Data quality assessment should also include examination of batch effects, technical artifacts, and missing data patterns. These issues can introduce biases that machine learning models will amplify instead of correct. Document all data processing steps so that results can be reproduced and audited.

Step 3: Evaluate Model Performance Metrics Appropriately

Model evaluation metrics require careful selection based on the task type. For clustering, classification, and regression tasks, different metrics apply, and each has advantages and disadvantages. Common pitfalls occur during model evaluation when researchers use inappropriate metrics for the task, fail to account for class imbalance, or evaluate on data that overlaps with training data. From a genomics perspective, researchers must understand how metric choice influences the critical interpretation of results.

For classification tasks in clinical applications, sensitivity and specificity matter differently depending on whether the cost of a false negative exceeds the cost of a false positive. For regression tasks in genomics, metrics that penalize large errors may be more appropriate than those that treat all errors equally. The choice of metric should follow from the biological or clinical consequence of different error types.

Step 4: Validate with Independent Data and Experimental Confirmation

Computational predictions require validation on independent datasets that were not used in model development. Cross-validation within a single dataset provides some protection against overfitting, but external validation on data from different sources or populations provides stronger evidence of generalizability. For predictions with therapeutic implications, experimental confirmation is essential.

The integration of bioinformatics and machine learning in cancer research illustrates this principle. In a study of baicalein against cervical cancer, bioinformatics and machine learning algorithms predicted six potential core targets. Molecular docking and molecular dynamics simulations validated these targets, with five showing significant binding affinity. PIM1 and CDK2 showed stable binding conformations in molecular dynamics simulations. Gene ontology and KEGG enrichment analyses indicated baicalein might regulate cell cycle progression via histone kinase mediated phosphorylation modifications. This multi-stage validation approach, moving from computational prediction to structural validation to pathway analysis, represents the standard for credible AI-driven discovery.

Step 5: Document Reproducibility and Version Control

Reproducibility requires more than saving the final analysis script. Document the software environment, package versions, random seeds, and parameter settings. Version control for both code and data ensures that analyses can be reconstructed exactly. For machine learning models, record the training and test splits, preprocessing steps, and any data transformations applied before model fitting.

The FAIR Guiding Principles provide a framework for making data findable, accessible, interoperable, and reusable. Applying these principles to research data and code supports reproducibility and enables other researchers to build on published work. For life sciences research, where datasets are often large and complex, FAIR practices reduce the friction of data sharing and increase the value of publicly funded research.

Step 6: Establish Escalation Criteria for Uncertain Results

Define in advance what will trigger additional validation, expert consultation, or abandonment of a computational approach. If model performance metrics fall below a predetermined threshold, if validation on independent data fails, or if results conflict with established biological knowledge, escalate the issue. For clinical applications, any prediction that would change patient management requires confirmation through established clinical pathways. For drug discovery, computational predictions of target engagement require experimental validation before advancing to preclinical studies.

Common Failure Patterns in AI-Driven Life Sciences Research

Overfitting to Training Data

The high-dimensional nature of genomic and clinical data makes overfitting a persistent risk. Models with many parameters can memorize training data instead of learn generalizable patterns. This manifests as excellent performance on training data with dramatic degradation on new data. Cross-validation helps detect overfitting, but only if the validation scheme respects the structure of the data, such as avoiding leakage between training and test sets when samples come from the same patient or the same batch.

Ignoring Batch Effects and Technical Artifacts

Biological data generated across different laboratories, sequencing platforms, or time periods contains technical variation unrelated to biology. Machine learning models will happily learn these artifacts if they correlate with the outcome of interest. This is particularly dangerous in multi-center studies where batch effects can create spurious associations. ComBat and similar tools can correct known batch effects, but unknown technical artifacts remain a challenge.

Data Leakage Between Training and Test Sets

Data leakage occurs when information from the test set influences model training. Common sources include normalizing the entire dataset before splitting, including duplicate samples across sets, or using features that are derived from the outcome. In genomics, leakage can occur when related samples, such as those from the same patient or family, appear in both training and test sets. This inflates performance estimates and leads to models that fail in real-world applications.

Overinterpreting Correlational Findings

Machine learning models identify correlations, not causes. A model that accurately predicts disease outcomes from gene expression data does not establish that the genes are causal. Causal machine learning approaches attempt to address this gap, but they require additional assumptions and often need perturbation data. Researchers should be explicit about whether their findings are predictive or causal and design follow-up experiments accordingly.

Poor Generalization Across Populations

Models trained on one population often perform poorly on others. This is particularly concerning in clinical applications where models developed on data from one demographic group may not generalize to other groups. The lung cancer machine learning literature explicitly identifies poor generalization across populations as a limitation. Researchers should evaluate model performance across relevant population subgroups and report performance stratified by these groups.

Emerging Trends and Future Directions

Causal Machine Learning

The shift from predictive to causal machine learning represents a significant trend in life sciences. Causal approaches aim to answer beyond what will happen but what would happen under intervention. In single-cell genomics, causal models can help identify which genes drive cellular transitions and which are merely correlated with them. This requires integrating perturbation data and making explicit assumptions about the underlying causal structure. The expectation is that causal models will become a crucial tool for informed experimental design as single-cell atlases and perturbation data become more available.

Multi-Omics Integration

Integrating data across genomics, transcriptomics, proteomics, metabolomics, and other omics layers provides a more complete picture of biological systems. Machine learning methods that can handle multiple data types simultaneously are emerging as powerful tools for understanding complex diseases. The integration of proteomic and transcriptomic data in acute myeloid leukemia research identified six genes whose expression predicted outcome, with IDH3B highly expressed in leukemic stem cells and associated with poor prognosis. This type of multi-omics integration reveals regulatory relationships that single-omics analyses miss.

Digital Twins and Multi-Scale Modeling

Predictive modeling in biology and medicine is moving toward digital twins and multi-scale modeling approaches. Digital twins are computational representations of biological systems that can simulate responses to interventions. Multi-scale models connect molecular, cellular, tissue, and organism-level processes. These approaches promise to accelerate drug development and enable personalized treatment planning, but they require extensive data and computational resources.

In Situ Structural Biology

The transition of cryo-EM from isolated proteins to in situ visualization of macromolecular architecture represents a major methodological advance. In situ single-particle methods collect high-dose, untilted images of cellular lamellae and use high-resolution templates for particle identification and refinement. These methods have significantly advanced both data throughput and achievable resolution, enabling structural biology in the native cellular context.

AI-Assisted Literature Synthesis

Artificial intelligence is being applied to the research literature itself. In breast cancer research, a combination of bioinformatic database analyses, artificial intelligence-assisted literature review, and manual literature review identified 15 biomarkers of clinical importance for CDK4/6 inhibitor response or resistance. This approach accelerates the synthesis of scattered findings into actionable knowledge, though it requires external expert validation to ensure accuracy and completeness.

Records, Measurements, and Quality Controls

Documentation Standards for AI Research

Maintain detailed records of data provenance, preprocessing steps, model parameters, and evaluation results. For each analysis, record the date, software versions, and computational environment. Store raw data separately from processed data and never overwrite raw files. Document any data exclusions and the rationale for them.

Quality Control Metrics

Establish quality control metrics appropriate to the data type. For sequencing data, track read depth, mapping rates, and duplication rates. For clinical data, monitor missingness patterns and outlier detection. For imaging data, document acquisition parameters and any artifacts. These metrics should be reported alongside analysis results so that reviewers can assess data quality.

Validation Logs

Maintain a validation log that records all attempts to validate model predictions, whether successful or not. This includes independent dataset validation, experimental confirmation, and clinical correlation. Negative validation results are as important as positive ones and should be documented to prevent others from repeating failed approaches.

Limitations and Professional Escalation Criteria

Known Limitations of Current Approaches

Machine learning in genetics and genomics carries the responsibility to exercise caution against biases and inflation of results. The gap between preclinical data and effective therapies in cancer research is attributed to suboptimal drug development for driver alterations, the high cost of clinical trials and available drugs, and limited access of patients to clinical trials. Bioinformatic analyses of complex data to characterize tumor biology, function, and dynamic tumor changes may improve cancer diagnosis, but harmonization between discoveries, policies, and practices is needed to expedite drug development.

For clinical decision support systems, the evidence base remains limited, particularly in specialized areas like pediatric oncology. Most applications concentrate in the treatment phase, with less emphasis on risk assessment, diagnosis, nursing workflow, and survivorship care. User interaction and follow-up functionalities require optimization, and multicenter randomized controlled trials are needed to validate clinical efficacy.

When to Escalate to Professional Consultation

Escalate to clinical or domain experts when computational results would inform patient care decisions, when predictions conflict with established biological knowledge, or when model performance falls below acceptable thresholds for the intended application. For drug discovery, escalate when computational predictions identify targets that would require significant resources to validate experimentally. For clinical applications, any prediction that would change patient management requires confirmation through established clinical pathways.

Safety and Regulatory Context

The application of AI in healthcare requires attention to regulatory standards and ethical considerations. The emphasis on integrating reliable and transparent AI solutions into clinical settings supports precision medicine, but regulatory standards for ethical and effective AI use in healthcare remain under development. Researchers should stay informed about evolving regulatory requirements and ensure that their work complies with applicable standards.

For genomic data, the Genomic Data Sharing Policy from the National Institutes of Health establishes expectations for responsible data sharing. Researchers should understand and comply with these requirements, including appropriate consent for data sharing and protection of participant privacy.

Frequently Asked Questions

What is the difference between bioinformatics and data science in life sciences?

Bioinformatics focuses on developing and applying computational methods for biological data, particularly sequence data, while data science in life sciences encompasses a broader range of methods including machine learning, statistical modeling, and data visualization applied to any biological or clinical data type. In practice, the fields overlap substantially, and many researchers work across both. Bioinformatics has traditionally emphasized database development, sequence alignment, and genome annotation, while data science brings in approaches from computer science and statistics for prediction and pattern discovery.

How do I choose between different machine learning methods for genomic data?

The choice depends on the task type, data characteristics, and interpretability requirements. For classification tasks with tabular genomic data, methods like random forest and support vector machines often perform well and offer some interpretability. For image data such as pathology slides or radiology scans, convolutional neural networks are typically the method of choice. For high-dimensional data with more features than samples, regularization methods or dimensionality reduction before modeling can help. Evaluate multiple methods using appropriate cross-validation and select based on performance on validation data, not training data.

What are the most common mistakes in evaluating machine learning models for genomics?

Common mistakes include using inappropriate metrics for the task, failing to account for class imbalance, evaluating on data that overlaps with training data, and ignoring batch effects. Data leakage between training and test sets is a frequent problem, particularly when related samples appear in both sets. Researchers should also be cautious about interpreting feature importance rankings, as correlated features can lead to unstable importance estimates.

How can I ensure my AI-based research is reproducible?

Use version control for code and data, document the computational environment including software versions and parameters, and record random seeds for any stochastic processes. Store raw data separately from processed data and document all preprocessing steps. Consider containerization tools to capture the full computational environment. Apply the FAIR Guiding Principles to make data findable, accessible, interoperable, and reusable.

What validation is needed before computational predictions can guide experiments?

Computational predictions should be validated on independent datasets that were not used in model development. For predictions with therapeutic implications, experimental confirmation is essential. The standard approach moves from computational prediction to structural validation to pathway analysis to experimental testing. For drug discovery, molecular docking and molecular dynamics simulations can validate binding affinity and stability before experimental testing.

How do I handle class imbalance in biological datasets?

Class imbalance occurs when one outcome class is rare, which is common in disease prediction where the number of cases is much smaller than controls. Options include resampling methods such as oversampling the minority class or undersampling the majority class, using class weights in model training, or choosing evaluation metrics that are appropriate for imbalanced data such as precision-recall curves instead of accuracy. The choice of approach depends on the biological question and the consequences of different error types.

What is causal machine learning and when should I use it?

Causal machine learning aims to identify cause-and-effect relationships instead of correlations. It is appropriate when the research question involves intervention, such as which gene to knock down or which drug to administer. Causal approaches require additional assumptions and often need perturbation data. In single-cell genomics, causal models can help identify which genes drive cellular transitions. These methods are more complex than predictive approaches and should be used when the research question genuinely requires causal answers.

How do I evaluate whether an AI tool is appropriate for clinical applications?

Clinical applications require evidence of efficacy from appropriate studies, ideally multicenter randomized controlled trials. The tool must demonstrate acceptable performance across relevant population subgroups, beyond in the population where it was developed. Interpretability is important for clinical acceptance, as clinicians need to understand the basis for recommendations. Integration with existing clinical workflows and electronic health records is also critical for practical adoption.

Related Bioinformatics Guides

• AI Protein Binder Design Tools: RFdiffusion, ProteinMPNN, BindCraft-Style Filtering, and Target-Specific Discovery Workflows
• Data Sharing and Privacy in Genomic Research
• Alternative Splicing Analysis from RNA-Seq Data
• Free Energy Perturbation Calculations in Drug Discovery
• The Discovery of the DNA Double Helix: A Data Story

References and Further Reading

• EMBL-EBI Training. European Bioinformatics Institute.
• NCBI Data Resources. National Center for Biotechnology Information.
• Genomic Data Sharing Policy. National Institutes of Health.
• The FAIR Guiding Principles. Scientific Data.
• Deciphering the mechanism of baicalein in cervical cancer via bioinformatics, machine learning and computational simulations: PIM1 and CDK2 are key target proteins.. International journal of biological macromolecules, 2025.
• Cyclic peptide structure prediction and design using AlphaFold2.. Nature communications, 2025.
• Identification and validation of SQLE in steroid-induced osteonecrosis of the femoral head: a bioinformatics and experimental study.. Journal of orthopaedic surgery and research, 2025.
• Screening the receptorome.. Journal of psychopharmacology (Oxford, England), 2006.
• Using bioinformatics and artificial intelligence to map the cyclin-dependent kinase 4/6 inhibitor biomarker landscape in breast cancer.. Future oncology (London, England), 2024.
• Targeted therapy in cancer.. Cancer chemotherapy and pharmacology, 2015.
• Understanding Membrane Protein Drug Targets in Computational Perspective.. Current drug targets, 2019.
• Proton-sensing G protein-coupled receptors: detectors of tumor acidosis and candidate drug targets.. Future medicinal chemistry, 2020.
• Integrating clinical decision support systems in pediatric oncology: A scoping review of applications, implementation gaps, and management Implications.. 2026.
• In situ single-particle Cryo-EM methods: Principle and applications.. 2026.
• TCA-controlled histone succinylation identifies IDH3B as a prognostic and therapeutic target in AML.. 2026.
• Complete genome sequence of Leuconostoc lactis B34-1-7-1 isolated from Panax ginseng in South Korea.. 2026.
• A Visualization Analysis of Machine Learning Applications in Gas Adsorption Using Nanoporous Materials.. 2026.
• Predictive modeling in biology and medicine: Digital twins and multi-scale modeling.. 2026.
• Machine Learning Frameworks for Prognosis and Detection of Lung Cancer: Combining Deep Learning, Genomics, and Radiomics. 2025 2nd International Conference on Software, Systems and Information Technology (SSITCON), 2025.
• Application of machine learning and genomics for orphan crop improvement. Nature Communications, 2025.
• Causal machine learning for single-cell genomics. Nature Genetics, 2025.
• A review of model evaluation metrics for machine learning in genetics and genomics. Frontiers Bioinform., 2024.
• Genomics, Phenomics, and Machine Learning in Transforming Plant Research: Advancements and Challenges. Horticultural Plant Journal, 2024.
• Causal machine learning for single-cell genomics. arXiv.org, 2023.
• Revolutionizing Healthcare: The Synergy of AI and Bioinformatics. Proceedings of IEEE International Conference on Signal Processing Computing and Control, 2025.
• Bioinformatics and beyond: AI applications in healthcare. Bioinformatics and Beyond AI Applications in Healthcare, 2025.
• AI in Bioinformatics and Genomics. International Conference on Intelligent and Innovative Practices in Engineering and Management 2024 Iipem 2024, 2024.
• Advancements in AI for Computational Biology and Bioinformatics: A Comprehensive Review. Methods in Molecular Biology, 2025.

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.