Open research questions in Bioinformatics and Genomic Networks

What the literature leaves open

• The Fisher metric becomes a liability when the distribution is a poor or misleading representation of the data, - Forcing an incorrect distributional assumption onto empirical data can distort downstream distances, - The one-gene approximation is limited when applied feature-by-feature to correlated expression

  When models choose metrics: Hidden geometry in computational biology · 2026 · DOI

• The gap between the choice of statistical model and the choice of distance metric. The lack of consideration for the geometry implied by a model when choosing a metric. The sensitivity of the Fisher-Rao metric to model misspecification.

  When models choose metrics: Hidden geometry in computational biology · 2026 · DOI

• Although high-throughput omics technologies have improved insight into disease mechanisms, data acquisition from inaccessible tissues such as the central nervous system remains a major limitation, causing small sample sizes and complicating early prediction of neurodegenerative disorders such as Alzheimer's and Parkinson's diseases.

  Adversarial random forests for omics synthesis · 2026 · DOI

• BRIDGE-AD supported an SPP1-centred cross-compartment hypothesis and nominated SCARB2, a poorly characterised candidate, for functional validation.

  BRIDGE-AD reveals Alzheimer's disease effectors through interpretable large-scale omics integration · 2026 · DOI

• The discovery of molecular relationships from high-dimensional data is a major open problem in bioinformatics.

  Interpretable AI for inference of causal molecular relationships from omics data · 2025 · DOI

• However, its hepatotoxic mechanisms remain insufficiently studied and systematically evaluated.

  Combined Analysis of Network Toxicology and Multiomics Revealed the Potential Mechanism of 6PPDQ-Induced Hepatotoxicity in Mice · 2025 · DOI

• Despite considerable research efforts on GBM, its underlying biological mechanisms remain unclear.

  Reversal gene expression assessment for drug repurposing, a case study of glioblastoma · 2025 · DOI

• The structural and functional effect of mutations in Trk-B & Trk-C proteins remains unclear.

  Unraveling the molecular mechanism of novel leukemia mutations on NTRK2 (A203T & R458G) and NTRK3 (E176D & L449F) genes using molecular dynamics simulations approach · 2024 · DOI

• The influence of replication timing on gene correlations remains unexplored. There is a need to integrate replication-timing proxies and morphological embeddings into individualized LIONESS gene networks.

  Personalized Morphology, Replication Timing, and RNA based Gene Expression Networks for Basal-like and Classical subtyping genes in Pancreatic Adenocarcinoma · 2026 · DOI

• Partial correlation networks were not measured at the patient level, but LIONESS networks were constructed [7]. The AUC is not strictly dependent on edge strength, but on which genes are included within modules influenced by morphological inputs. Increasing the number of nodes or genes lowers network resolution. STRING database connections were attempted; however, connections among many Moffitt genes were incomplete or absent, which is expected for sparse or poorly annotated regulatory interactions [8,9]. The globally constructed net- works are not subtype-specific at the patient level but can characterize cohort-level variance related to tumor progression. Nodes were not constructed from morphology and replication timing directly; instead, these modalities influenced edge weights and module scoring.

  Personalized Morphology, Replication Timing, and RNA based Gene Expression Networks for Basal-like and Classical subtyping genes in Pancreatic Adenocarcinoma · 2026 · DOI

• Exploring the potential of multimodal data fusion to improve prediction accuracy. Investigating cross-species prediction of essential proteins to improve the model's generalization ability.

  A CSGNN model-based method for essential protein identification · 2026 · DOI

• The ablation study section is incomplete in the excerpt (ends with 'The full CSGNN m'), making a comprehensive assessment of individual component contributions impossible.

  A CSGNN model-based method for essential protein identification · 2026 · DOI

• Existing approaches for integrating heterogeneous multi-omics data and modeling gene-gene interactions often face challenges. Current diagnostic methods have limitations, including the inability to differentiate between primary cancer and metastatic tumors.

  SEMO-GCN: Semantic Enhanced Multi-Omics Graph Representation Learning for Pan-Cancer Metastasis Identification · 2026 · DOI

• Metastasis of cancer remains a significant challenge in oncology and accounts for the majority of cancer-related fatalities globally. Traditional methods for predicting metastasis often fail to incorporate complex multi-omics data and to understand the intricate molecular interactions that promote tumor growth. In this study, we created a new framework that combines multi-omics data like mRNA expression, DNA methylation, gene mutation, and copy number alteration (CNA) with Graph Convolutional Networks (GCNs) and gene embeddings from large language models (LLM) to make pan-cancer metastasis prediction integrates more molecular-level omics characteristics with LLM- derived semantic representations of genes. This lets it topological and contextual biological see both proposed model accurate.The Page | 29 Journal of Hunan University (Natural Sciences） Vol. 53 No. 2, February 2026 relationships, which gives a fuller picture of how tumors act. The proposed graph-based GCN (Omics + LLM) model is a lot better than all of the non-graph baselines, like Transformer, MLP, and Random Forest, when you compare them. It has a 97.37% accuracy, a 97.06% F1- score, and a 99.72% AUC, which is a big improvement over older methods. This shows that adding CNA features to other omics types, as well as PPI- network topology and LLM-based embeddings, makes it easier to understand and predict biological processes. Adding more omics types (like proteomics and metabolomics) and real-time clinical data to this framework will make it better in the future.This will help find problems early and give each patient the best care.Another important goal will be to make it easier to apply what we’ve learned to different datasets and groups of people. Finally, we’ll look into how to use biomedical LLMs with attention and reinforcement learning systems to model how genes interact with each other over time and make predictions about metastasis that are even more useful in the clinic.

  SEMO-GCN: Semantic Enhanced Multi-Omics Graph Representation Learning for Pan-Cancer Metastasis Identification · 2026 · DOI

• The proteome-wide computational evaluation of the effects of alternative splicing on protein interaction networks remains challenging. The lack of extensive experimental evidence makes it difficult to understand the large-scale functional regulation details through AS.

  DeepISO: deep learning-powered prediction of protein–protein interaction rewiring generated by alternative splicing · 2026 · DOI

• The creation of GO subset files can be time-and labor-intensive. The integration of non-plant GO terms into the GO Big subsets. The need for a balance between functional specificity and biological relevance.

  Go Big or go home: a new gene ontology subset that improves plant gene function prediction · 2026 · DOI

• The application of the GO Big subsets to other species. The development of new methods for creating GO subsets. The integration of the GO Big subsets with other gene function prediction tools.

  Go Big or go home: a new gene ontology subset that improves plant gene function prediction · 2026 · DOI

• Integrating functional dependency data into the framework. Prospective experimental validation of the framework's predictions. Adapting the framework to diverse cancer types and complex diseases.

  Commentary: GETgene-AI: a framework for prioritizing actionable cancer drug targets · 2026 · DOI

• The lack of a scalable, evidence-based, and interpretable platform for target prioritization. The need for a framework that can adapt to diverse cancer types and complex diseases. The limitations of conventional unidimensional analyses in prioritizing therapeutically actionable targets.

  Commentary: GETgene-AI: a framework for prioritizing actionable cancer drug targets · 2026 · DOI

• The study has limitations in terms of computational feasibility and the need for further validation. The framework may not be applicable to all types of diseases or therapeutic areas.

  A biology-based quality-diversity algorithm for drug repurposing in Alzheimer’s disease using automated machine learning · 2026 · DOI

• Further validation of the identified candidates. Application of the framework to other diseases or therapeutic areas. Investigation of the potential synergistic effects of the identified candidates.

  A biology-based quality-diversity algorithm for drug repurposing in Alzheimer’s disease using automated machine learning · 2026 · DOI

• The study faced the challenge of identifying functionally coherent transcript modules in non-model plant species. The approach needed to overcome the limitation of relying primarily on expression-level comparisons.

  Annotation-Based Clustering Reveals Functional Transcript Modules Involved in Stress Response and Metabolic Pathways in Cinnamomum · 2026 · DOI

• Future studies can use this approach to identify functionally coherent transcript modules in other non-model plant species. The findings can inform strategies for improving plant stress response and metabolic pathways.

  Annotation-Based Clustering Reveals Functional Transcript Modules Involved in Stress Response and Metabolic Pathways in Cinnamomum · 2026 · DOI

• Further study of the role of PAI-1 in various diseases, including cancer and neurodegeneration. Investigation of the potential therapeutic benefits of the use of aspirin and statins in these diseases. Development of combination therapies, including the modulation of PAI-1, for the treatment of these diseases.

  PAI-1 as a Universal Systemic Integrator Node: A Unified Relative Permissiveness Network Synthesis Across Cancer, Neurodegeneration, Acute Infection, Fibrosis, and Immune Biology · 2026 · DOI

• The current lack of effective therapies for chronic diseases, such as cancer and neurodegeneration. The need for a new approach to treating these diseases, focusing on the modulation of key players such as PAI-1.

  PAI-1 as a Universal Systemic Integrator Node: A Unified Relative Permissiveness Network Synthesis Across Cancer, Neurodegeneration, Acute Infection, Fibrosis, and Immune Biology · 2026 · DOI

Most-cited papers in Bioinformatics and Genomic Networks

• Tissue-based map of the human proteome · Science · 2015 · 14,290 citations
• DAVID: a web server for functional enrichment analysis and functional annotation of gene lists (2021 update) · Nucleic Acids Research · 2022 · 6,077 citations
• The Connectivity Map: Using Gene-Expression Signatures to Connect Small Molecules, Genes, and Disease · Science · 2006 · 4,683 citations
• Detecting Novel Associations in Large Data Sets · Science · 2011 · 2,928 citations
• Transfer learning enables predictions in network biology · Nature · 2023 · 1,212 citations
• decoupleR: ensemble of computational methods to infer biological activities from omics data · Bioinformatics Advances · 2022 · 1,127 citations
• Personalized medicine: motivation, challenges, and progress · Fertility and Sterility · 2018 · 871 citations
• Majorbio Cloud: A one‐stop, comprehensive bioinformatic platform for multiomics analyses · iMeta · 2022 · 851 citations
• Differential network biology · Molecular Systems Biology · 2012 · 683 citations
• OpenCell: Endogenous tagging for the cartography of human cellular organization · Science · 2022 · 583 citations

Most recent work

• A complete map of specificity encoding enables reprogramming of a protein interaction · bioRxiv · 2026
• Rethinking network analysis in ethnopharmacology: a multi-omics and AI roadmap to overcome conceptual and methodological biases · Frontiers in Pharmacology · 2026
• Deep Learning-Enabled Multi-Omics Integration: A New Frontier in Precise Drug Target Discovery · Biology · 2026
• Medea: An AI agent for therapeutic reasoning across biological contexts · bioRxiv · 2026
• Global genetic interaction network of a human cell maps conserved principles and informs functional interpretation of gene co-essentiality profiles · Cell · 2026
• FlyPredictome: A structural atlas of predicted protein-protein interactions in Drosophila · bioRxiv · 2026
• AggrescanAI: Prediction of Aggregation-Prone Regions Using Contextualized Embeddings · Journal of Molecular Biology · 2026
• Expanding the scope of protein language modeling to protein-protein interactions with MSA Pairformer · Cell · 2026
• Biologically informed neural network models are robust to spurious interactions via self-pruning · bioRxiv · 2026
• Lemonite: identification of regulatory metabolites through data-driven, interpretable integration of transcriptomics and metabolomics data · bioRxiv · 2026

Related topics in Biochemistry, Genetics and Molecular Biology