Gene expression regulation underpins cellular function and disease progression, yet its complexity and the limitations of conventional detection methods hinder clinical translation. In this review, we define “predict” as the AI-driven inference of gene expression levels and regulatory mechanisms from non-invasive multimodal data (e.g., histopathology images, genomic sequences, and electronic health records) instead of direct molecular assays. We systematically examine and analyze the current approaches for predicting gene expression and diagnosing diseases, highlighting their respective advantages and limitations. Machine learning algorithms and deep learning models excel in extracting meaningful features from diverse biomedical modalities, enabling tools like PathChat and Prov-GigaPath to improve cancer subtyping, therapy response prediction, and biomarker discovery. Despite significant progress, persistent challenges—such as data heterogeneity, noise, and ethical issues including privacy and algorithmic bias—still limit broad clinical adoption. Emerging solutions like cross-modal pretraining frameworks, federated learning, and fairness-aware model design aim to overcome these barriers. Case studies in precision oncology illustrate AI’s ability to decode tumor ecosystems and predict treatment outcomes. By harmonizing multimodal data and advancing ethical AI practices, this field holds immense potential to propel personalized medicine forward, although further innovation is needed to address the issues of scalability, interpretability, and equitable deployment.

The low capture rate of expressed RNAs from single-cell sequencing technology is one of the major obstacles to downstream functional genomics analyses. Recently, a number of imputation methods have emerged for single-cell transcriptome data, however, recovering missing values in very sparse expression matrices remains a substantial challenge. Here, we propose a new algorithm, WEDGE (WEighted Decomposition of Gene Expression), to impute gene expression matrices by using a biased low-rank matrix decomposition method (bLRMD). WEDGE successfully recovered expression matrices, reproduced the cell-wise and gene-wise correlations, and improved the clustering of cells, performing impressively for applications with multiple cell type datasets with high dropout rates. Overall, this study demonstrates a potent approach for imputing sparse expression matrix data, and our WEDGE algorithm should help many researchers to more profitably explore the biological meanings embedded in their scRNA-seq datasets.

The skin acts as the first barrier protecting the human body against the external environment. Transcriptional heterogeneity in human skin is widely confirmed, but the regulatory mechanisms remain largely unexplored. In this study, we employed high-throughput single-cell chromatin accessibility and transcriptome sequencing (HT-scCAT-seq), a technique for simultaneous analysis of the transcriptome and epigenome. We used HT-scCAT-seq to analyze 10,065 cell profiles from four adult human eyelid skin and define 13 distinct cell types. In addition, we described detailed molecular signatures and identified key gene regulatory network enriched in each cell type. Our dataset is a valuable resource for further research in human skin biology.