The promise of explainable deep learning for omics data analysis: Adding new discovery tools to AI

AI and omics technologies in biobanking: Applications and challenges for public health

OBJECTIVES

Considering the growing intersection of biobanks, artificial intelligence (AI) and omics research, and their critical impact on public health, this study aimed to explore the current and future public health implications and challenges of AI and omics-driven innovations in biobanking.

STUDY DESIGN

Narrative literature review.

METHODS

A structured literature search was conducted in Scopus, PubMed, Web of Science and IEEExplore databases using relevant search terms. Additional references were identified through backward and forward citation chaining. Key themes were aggregated and analysed through thematic analysis.

RESULTS

Thirty-seven studies were selected for analysis, leading to the identification and categorisation of key developments. Several key technical, ethical and implementation challenges were also identified, including AI model selection, data accessibility, variability and quality issues, lack of robust and standardised validation methods, explainability, accountability, lack of transparency, algorithmic bias, privacy, security and fairness issues, and governance model selection. Based on these results, potential future scenarios of AI and omics integration in biobanking and their related public health implications were considered.

CONCLUSIONS

While AI and omics-driven innovations in biobanking offer specific transformative public health benefits, addressing their technical, ethical and implementation challenges is crucial. Robust regulatory frameworks, feasible governance models, access to quality data, interdisciplinary collaboration, and transparent and validated AI systems are essential to maximise benefits and mitigate risks. Further research and policy development are needed to support the responsible integration of these technologies in biobanking and public health.

Microbial Ecotoxicology-40 Years on

Although ecotoxicology was emerging as a field through the 1970s, the incorporation of microbial indicators into the framework has been slower to evolve. The exploration of microbes as sensitive toxicity tests began in the late 70s and early 80s (with the emergence of Microtox® and other simple tests). However, the applications have been limited, beyond water and wastewater screening. This opinion piece reflects my own perspective on the field-from my early excitement in the 1990s for its possibilities, to a sense of frustration at the slow pace of new development and applications in the field-despite the surge of "omics" options. While microbiology still fails to lead the field of ecotoxicology, the potential remains.

Graph-Based 3-Dimensional Spatial Gene Neighborhood Networks of Single Cells in Gels and Tissues

Objective: We developed 3-dimensional spatially resolved gene neighborhood network embedding (3D-spaGNN-E) to find subcellular gene proximity relationships and identify key subcellular motifs in cell-cell communication (CCC). Impact Statement: The pipeline combines 3D imaging-based spatial transcriptomics and graph-based deep learning to identify subcellular motifs. Introduction: Advancements in imaging and experimental technology allow the study of 3D spatially resolved transcriptomics and capture better spatial context than approximating the samples as 2D. However, the third spatial dimension increases the data complexity and requires new analyses. Methods: 3D-spaGNN-E detects single transcripts in 3D cell culture samples and identifies subcellular gene proximity relationships. Then, a graph autoencoder projects the gene proximity relationships into a latent space. We then applied explainability analysis to identify subcellular CCC motifs. Results: We first applied the pipeline to mesenchymal stem cells (MSCs) cultured in hydrogel. After clustering the cells based on the RNA count, we identified cells belonging to the same cluster as homotypic and those belonging to different clusters as heterotypic. We identified changes in local gene proximity near the border between homotypic and heterotypic cells. When applying the pipeline to the MSC-peripheral blood mononuclear cell (PBMC) coculture system, we identified CD4+ and CD8+ T cells. Local gene proximity and autoencoder embedding changes can distinguish strong and weak suppression of different immune cells. Lastly, we compared astrocyte-neuron CCC in mouse hypothalamus and cortex by analyzing 3D multiplexed-error-robust fluorescence in situ hybridization (MERFISH) data and identified regional gene proximity differences. Conclusion: 3D-spaGNN-E distinguished distinct CCCs in cell culture and tissue by examining subcellular motifs.

Early and noninvasive prediction of response to neoadjuvant therapy for breast cancer via longitudinal ultrasound and MR deep learning: A multicentre study

RATIONALE AND OBJECTIVES

The early prediction of response to neoadjuvant chemotherapy (NAC) will aid in the development of personalized treatments for patients with breast cancer. This study investigated the value of longitudinal multimodal deep learning (DL) based on breast MR and ultrasound (US) in predicting pathological complete response (pCR) after NAC.

MATERIALS AND METHODS

We retrospectively reviewed the pre-NAC and post-2nd-NAC MR and/or US images of 448 patients enrolled from three centers and extracted DL features from the largest section of the breast tumour using ResNet50. T test, Pearson correlation analysis and least absolute shrinkage and selection operator regression were used to select the most significant DL features for the pre-NAC and post-2nd-NAC MR and US DL models. The stacking model integrates different single-modality DL models and meaningful clinical data. The diagnostic performance of the models was evaluated.

RESULTS

In all the patients, the pCR rate was 36.65%. There was no significant difference in diagnostic performance between the different single-modality DL models (DeLong test, p > 0.05). The stacking model integrating the above four DL models with HER2 status yielded areas under the curves of 0.951-0.979, accuracies of 91.55%-92.65%, sensitivities of 90.63%-93.94%, and specificities of 89.47%-94.44% in the cohorts.

CONCLUSION

Longitudinal multimodal DL can be useful in predicting pCR. The stacking model can be used as a new tool for the early noninvasive prediction of the response to NAC, as evidenced by its excellent performance, and therefore aid the development of personalized treatment strategies for patients with breast cancer.

Interpretable and explainable predictive machine learning models for data-driven protein engineering

Protein engineering through directed evolution and (semi)rational design has become a powerful approach for optimizing and enhancing proteins with desired properties. The integration of artificial intelligence methods has further accelerated protein engineering process by enabling the development of predictive models based on data-driven strategies. However, the lack of interpretability and transparency in these models limits their trustworthiness and applicability in real-world scenarios. Explainable Artificial Intelligence addresses these challenges by providing insights into the decision-making processes of machine learning models, enhancing their reliability and interpretability. Explainable strategies has been successfully applied in various biotechnology fields, including drug discovery, genomics, and medicine, yet its application in protein engineering remains underexplored. The incorporation of explainable strategies in protein engineering holds significant potential, as it can guide protein design by revealing how predictive models function, benefiting approaches such as machine learning-assisted directed evolution. This perspective work explores the principles and methodologies of explainable artificial intelligence, highlighting its relevance in biotechnology and its potential to enhance protein design. Additionally, three theoretical pipelines integrating predictive models with explainable strategies are proposed, focusing on their advantages, disadvantages, and technical requirements. Finally, the remaining challenges of explainable artificial intelligence in protein engineering and future directions for its development as a support tool for traditional protein engineering methodologies are discussed.

Delta dual‑region DCE-MRI radiomics from breast masses predicts axillary lymph node response after neoadjuvant therapy for breast cancer

OBJECTIVES

This study was designed to develop and validate models based on delta intratumoral and peritumoral radiomics features from breast masses on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) for the prediction of axillary lymph node (ALN) pathological complete response (pCR) after neoadjuvant therapy (NAT) in patients with breast cancer (BC).

METHODS

We retrospectively collected data from 187 BC patients with ALN metastases. Radiomics features were extracted from the intratumoral and 3 mm-peritumoral regions on DCE-MRI at baseline and after the 2nd course of NAT to calculate delta intratumoral and peritumoral radiomics features, respectively. After feature selection, the delta intratumoral radiomics (DIR) model and delta peritumoral radiomics (DPR) model were built using the retained features. An ultrasound model was constructed on the basis of preoperative axillary ultrasound results. All variables were screened by univariate and multivariate logistic regression to construct the combined model. The above models were evaluated and compared.

RESULTS

In the validation set, the ultrasound model had the lowest AUC, which was lower than those of the DIR, DPR and combined models (0.627 vs 0.825, 0.687, 0.846, respectively). The combined model constructed by delta dual-region radiomics and ultrasound dianogsis was significantly better than the ultrasound model in terms of the Delong test and integrated discrimination improvement (all p < 0.05).

CONCLUSIONS

Delta intratumoral and peritumoral radiomics based on DCE-MRI have the potential to predict ALN status after NAT. The combined model based on delta dual-region radiomics of breast mass can accurately diagnose ALN-pCR and provide assistance in the selection of axillary surgical approaches for patients.

Opportunities, challenges and future perspectives of using bioinformatics and artificial intelligence techniques on tropical disease identification using omics data

Tropical diseases can often be caused by viruses, bacteria, parasites, and fungi. They can be spread over vectors. Analysis of multiple omics data types can be utilized in providing comprehensive insights into biological system functions and disease progression. To this end, bioinformatics tools and diverse AI techniques are pivotal in identifying and understanding tropical diseases through the analysis of omics data. In this article, we provide a thorough review of opportunities, challenges, and future directions of utilizing Bioinformatics tools and AI-assisted models on tropical disease identification using various omics data types. We conducted the review from 2015 to 2024 considering reliable databases of peer-reviewed journals and conference articles. Several keywords were taken for the article searching and around 40 articles were reviewed. According to the review, we observed that utilization of omics data with Bioinformatics tools like BLAST, and Clustal Omega can make significant outcomes in tropical disease identification. Further, the integration of multiple omics data improves biomarker identification, and disease predictions including disease outbreak predictions. Moreover, AI-assisted models can improve the precision, cost-effectiveness, and efficiency of CRISPR-based gene editing, optimizing gRNA design, and supporting advanced genetic correction. Several AI-assisted models including XAI can be used to identify diseases and repurpose therapeutic targets and biomarkers efficiently. Furthermore, recent advancements including Transformer-based models such as BERT and GPT-4, have been mainly applied for sequence analysis and functional genomics. Finally, the most recent GeneViT model, utilizing Vision Transformers, and other AI techniques like Generative Adversarial Networks, Federated Learning, Transfer Learning, Reinforcement Learning, Automated ML and Attention Mechanism have shown significant performance in disease classification using omics data.

Progress in the application of artificial intelligence in skin wound assessment and prediction of healing time

Since the 1970s, artificial intelligence (AI) has played an increasingly pivotal role in the medical field, enhancing the efficiency of disease diagnosis and treatment. Amidst an aging population and the proliferation of chronic disease, the prevalence of complex surgeries for high-risk multimorbid patients and hard-to-heal wounds has escalated. Healthcare professionals face the challenge of delivering safe and effective care to all patients concurrently. Inadequate management of skin wounds exacerbates the risk of infection and complications, which can obstruct the healing process and diminish patients' quality of life. AI shows substantial promise in revolutionizing wound care and management, thus enhancing the treatment of hospitalized patients and enabling healthcare workers to allocate their time more effectively. This review details the advancements in applying AI for skin wound assessment and the prediction of healing timelines. It emphasizes the use of diverse algorithms to automate and streamline the measurement, classification, and identification of chronic wound healing stages, and to predict wound healing times. Moreover, the review addresses existing limitations and explores future directions.

Bioinformatics and machine learning to support nanomaterial grouping

Nanomaterials (NMs) offer plenty of novel functionalities. Moreover, their physicochemical properties can be fine-tuned to meet the needs of specific applications, leading to virtually unlimited numbers of NM variants. Hence, efficient hazard and risk assessment strategies building on New Approach Methodologies (NAMs) become indispensable. Indeed, the design, the development and implementation of NAMs has been a major topic in a substantial number of research projects. One of the promising strategies that can help to deal with the high number of NMs variants is grouping and read-across. Based on demonstrated structural and physicochemical similarity, NMs can be grouped and assessed together. Within an established NM group, read-across may be performed to fill in data gaps for data-poor variants using existing data for NMs within the group. Establishing a group requires a sound justification, usually based on a grouping hypothesis that links specific physicochemical properties to well-defined hazard endpoints. However, for NMs these interrelationships are only beginning to be understood. The aim of this review is to demonstrate the power of bioinformatics with a specific focus on Machine Learning (ML) approaches to unravel the NM Modes-of-Action (MoA) and identify the properties that are relevant to specific hazards, in support of grouping strategies. This review emphasizes the following messages: 1) ML supports identification of the most relevant properties contributing to specific hazards; 2) ML supports analysis of large omics datasets and identification of MoA patterns in support of hypothesis formulation in grouping approaches; 3) omics approaches are useful for shifting away from consideration of single endpoints towards a more mechanistic understanding across multiple endpoints gained from one experiment; and 4) approaches from other fields of Artificial Intelligence (AI) like Natural Language Processing or image analysis may support automated extraction and interlinkage of information related to NM toxicity. Here, existing ML models for predicting NM toxicity and for analyzing omics data in support of NM grouping are reviewed. Various challenges related to building robust models in the field of nanotoxicology exist and are also discussed.

Reviewing the essential roles of remote phenotyping, GWAS and explainable AI in practical marker-assisted selection for drought-tolerant winter wheat breeding

Marker-assisted selection (MAS) plays a crucial role in crop breeding improving the speed and precision of conventional breeding programmes by quickly and reliably identifying and selecting plants with desired traits. However, the efficacy of MAS depends on several prerequisites, with precise phenotyping being a key aspect of any plant breeding programme. Recent advancements in high-throughput remote phenotyping, facilitated by unmanned aerial vehicles coupled to machine learning, offer a non-destructive and efficient alternative to traditional, time-consuming, and labour-intensive methods. Furthermore, MAS relies on knowledge of marker-trait associations, commonly obtained through genome-wide association studies (GWAS), to understand complex traits such as drought tolerance, including yield components and phenology. However, GWAS has limitations that artificial intelligence (AI) has been shown to partially overcome. Additionally, AI and its explainable variants, which ensure transparency and interpretability, are increasingly being used as recognised problem-solving tools throughout the breeding process. Given these rapid technological advancements, this review provides an overview of state-of-the-art methods and processes underlying each MAS, from phenotyping, genotyping and association analyses to the integration of explainable AI along the entire workflow. In this context, we specifically address the challenges and importance of breeding winter wheat for greater drought tolerance with stable yields, as regional droughts during critical developmental stages pose a threat to winter wheat production. Finally, we explore the transition from scientific progress to practical implementation and discuss ways to bridge the gap between cutting-edge developments and breeders, expediting MAS-based winter wheat breeding for drought tolerance.

The future of equine semen analysis

We are currently experiencing a period of rapid advancement in various areas of science and technology. The integration of high throughput 'omics' techniques with advanced biostatistics, and the help of artificial intelligence, is significantly impacting our understanding of sperm biology. These advances will have an appreciable impact on the practice of reproductive medicine in horses. This article provides a brief overview of recent advances in the field of spermatology and how they are changing assessment of sperm quality. This article is written from the authors' perspective, using the stallion as a model. We aim to portray a brief overview of the changes occurring in the assessment of sperm motility and kinematics, advances in flow cytometry, implementation of 'omics' technologies, and the use of artificial intelligence/self-learning in data analysis. We also briefly discuss how some of the advances can be readily available to the practitioner, through the implementation of 'on-farm' devices and telemedicine.

Boosting grapevine breeding for climate-smart viticulture: from genetic resources to predictive genomics

The multifaceted nature of climate change is increasing the urgency to select resilient grapevine varieties, or generate new, fitter cultivars, to withstand a multitude of new challenging conditions. The attainment of this goal is hindered by the limiting pace of traditional breeding approaches, which require decades to result in new selections. On the other hand, marker-assisted breeding has proved useful when it comes to traits governed by one or few genes with great effects on the phenotype, but its efficacy is still restricted for complex traits controlled by many loci. On these premises, innovative strategies are emerging which could help guide selection, taking advantage of the genetic diversity within the Vitis genus in its entirety. Multiple germplasm collections are also available as a source of genetic material for the introgression of alleles of interest via adapted and pioneering transformation protocols, which present themselves as promising tools for future applications on a notably recalcitrant species such as grapevine. Genome editing intersects both these strategies, not only by being an alternative to obtain focused changes in a relatively rapid way, but also by supporting a fine-tuning of new genotypes developed with other methods. A review on the state of the art concerning the available genetic resources and the possibilities of use of innovative techniques in aid of selection is presented here to support the production of climate-smart grapevine genotypes.

Proteomic approaches advancing targeted protein degradation

Targeted protein degradation (TPD) is an emerging modality for research and therapeutics. Most TPD approaches harness cellular ubiquitin-dependent proteolytic pathways. Proteolysis-targeting chimeras (PROTACs) and molecular glue (MG) degraders (MGDs) represent the most advanced TPD approaches, with some already used in clinical settings. Despite these advances, TPD still faces many challenges, pertaining to both the development of effective, selective, and tissue-penetrant degraders and understanding their mode of action. In this review, we focus on progress made in addressing these challenges. In particular, we discuss the utility and application of recent proteomic approaches as indispensable tools to enable insights into degrader development, including target engagement, degradation selectivity, efficacy, safety, and mode of action.

Multimodal deep learning as a next challenge in nutrition research: tailoring fermented dairy products based on cytidine diphosphate-diacylglycerol synthase-mediated lipid metabolism

Deep learning is evolving in nutritional epidemiology to address challenges including precise nutrition and data-driven disease modeling. Fermented dairy products consumption as the implementation of specific dietary priority contributes to a lower risk of all-cause mortality, cardiovascular disease, and obesity. Various lipid types play different roles in cardiometabolic health and fermentation process changes the lipid profile in dairy products. Leveraging the power of multiple biological datasets can provide mechanistic insights into how proteins impact lipid pathways, and establish connections among fermentation-lipid biomarkers-protein. The recent leap of deep learning has been performed in food category recognition, agro-food freshness detection, and food flavor prediction and regulation. The proposed multimodal deep learning method includes four steps: (i) Forming data matrices based on data generated from different omics layers. (ii) Decomposing high-dimensional omics data according to self-attention mechanism. (iii) Constructing View Correlation Discovery Network to learn the cross-omics correlations and integrate different omics datasets. (iv) Depicting a biological network for lipid metabolism-centered quantitative multi-omics data analysis. Relying on the cytidine diphosphate-diacylglycerol synthase-mediated lipid metabolism regulates the glycerophospholipid composition of fermented dairy effectively. Innovative processing strategies including ohmic heating and pulsed electric field improve the sensory qualities and nutritional characteristics of the products.