Jianxiao Liu

Chromatin interactions establish spatial proximity between distant regulatory elements and their target genes, significantly influencing gene expression, and phenotypic traits. In this study, we present a plant chromatin interaction prediction model called PlantCTCIP based on Convolutional Neural Networks and Transformer. PlantCTCIP demonstrated superior performance compared to the conventional models. Specifically, PlantCTCIP improved the average AUC of chromatin interaction predictions by 14.56% across the four species in PPI (proximal promoter interaction) mode. Similarly, PlantCTCIP improved the average AUC of chromatin interaction predictions by 9.6% in the PDI (distal promoter interaction) mode. We constructed genome‐wide chromatin interaction maps for four plants (maize, rice, cotton and wheat) through PlantCTCIP, further used the Hi‐C experiment to validate correctness of the predicted PPIs and PDIs. Some key motifs that influence chromatin interactions are identified, and they are significantly enriched in expression quantitative trait loci (eQTLs) and open chromatin regions. We also analysed the enrichment and species specificity of the transcription factors (TF) and synergistic network of TFs that affect PPIs and PDIs of four crops. Using cloned genes ( ZmRAVL1 , ZmRPG , ZmRap2.7 and GaFZ ) of maize and cotton as examples, PlantCTCIP can assist in identifying target genes regulated by distal elements and mining functional sites combined with chromatin conformation capture (3C) experiments. This research helps to analyse the regulatory mechanism of gene expression and provides novel perspectives for intelligent design breeding of diverse crops. PlantCTCIP is available at http://www.plantctcip.com .

SUMMARY: Due to its excellent performance in processing large amounts of data and capturing complex non‐linear relationships, deep learning has been widely applied in many fields of plant biology. Here we first review the application of deep learning in analyzing genome sequences to predict gene expression, chromatin interactions, and epigenetic features (open chromatin, transcription factor binding sites, and methylation sites) in plants. Then, current motif mining and functional component design and synthesis based on generative adversarial networks, large models, and attention mechanisms are elaborated in detail. The progress of protein structure and function prediction, genomic prediction, and large model applications based on deep learning is also discussed. Finally, this work provides prospects for the future development of deep learning in plants with regard to multiple omics data, algorithm optimization, large language models, sequence design, and intelligent breeding.

Detecting genes that affect specific traits (such as human diseases and crop yields) is important for treating complex diseases and improving crop quality. A genome-wide association study (GWAS) provides new insights and directions for understanding complex traits by identifying important single nucleotide polymorphisms. Many GWAS summary statistics data related to various complex traits have been gathered recently. Studies have shown that GWAS risk loci and expression quantitative trait loci (eQTLs) often have a lot of overlaps, which makes gene expression gradually become an important intermediary to reveal the regulatory role of GWAS. In this review, we review three types of gene-trait association detection methods of integrating GWAS summary statistics and eQTLs data, namely colocalization methods, transcriptome-wide association study-oriented approaches, and Mendelian randomization-related methods. At the theoretical level, we discussed the differences, relationships, advantages, and disadvantages of various algorithms in the three kinds of gene-trait association detection methods. To further discuss the performance of various methods, we summarize the significant gene sets that influence high-density lipoprotein, low-density lipoprotein, total cholesterol, and triglyceride reported in 16 studies. We discuss the performance of various algorithms using the datasets of the four lipid traits. The advantages and limitations of various algorithms are analyzed based on experimental results, and we suggest directions for follow-up studies on detecting gene-trait associations.

Mining the epistatic gene loci for complex diseases is an important research topic in recent years. The existing epistasis detection methods have the shortcomings of high complexity, low efficiency, high false positive rate, inability to deal with large-scale genome-wide data, etc. In this work, we propose a epistasis mining method which using k-tree optimizing Bayesian network (BN) (KtreeBnIp). Firstly, it constructs the k-tree including large-scale of SNP loci and phenotype by sampling the Dandelion code uniformly. Then it decomposes k-tree into different k-cliques by using the graph decomposition algorithm based on neighbor nodes. For the nodes in each k-clique, the integer linear programming optimizing Bayesian network (ILPBN) is used to obtain the sub-network quickly and accurately. Finally, it merges all the sub-networks to obtain the whole network. It repeats the above operations (k-tree generation, k-tree decomposition, network generation) several times and gets the final network according to the frequency of edges, and thus to obtain the epistatic loci affecting phenotype. We compare KtreeBnIp with the current popular epistasis mining algorithms using both simulated and real age-related macular degeneration(AMD) datasets. Experiment results show that KtreeBnIp has better epistasis detection accuracy, higher F1-score, and lower false positive rate. Most importantly, it can be used in large-scale of SNPs for epistasis detection.

Identifying the function of DNA sequences accurately is an essential and challenging task in the genomic field. Until now, deep learning has been widely used in the functional analysis of DNA sequences, including DeepSEA, DanQ, DeepATT and TBiNet. However, these methods have the problems of high computational complexity and not fully considering the distant interactions among chromatin features, thus affecting the prediction accuracy. In this work, we propose a hybrid deep neural network model, called DeepFormer, based on convolutional neural network (CNN) and flow-attention mechanism for DNA sequence function prediction. In DeepFormer, the CNN is used to capture the local features of DNA sequences as well as important motifs. Based on the conservation law of flow network, the flow-attention mechanism can capture more distal interactions among sequence features with linear time complexity. We compare DeepFormer with the above four kinds of classical methods using the commonly used dataset of 919 chromatin features of nearly 4.9 million noncoding DNA sequences. Experimental results show that DeepFormer significantly outperforms four kinds of methods, with an average recall rate at least 7.058% higher than other methods. Furthermore, we confirmed the effectiveness of DeepFormer in capturing functional variation using Alzheimer’s disease, pathogenic mutations in alpha-thalassemia and modification in CCCTC-binding factor (CTCF) activity. We further predicted the maize chromatin accessibility of five tissues and validated the generalization of DeepFormer. The average recall rate of DeepFormer exceeds the classical methods by at least 1.54%, demonstrating strong robustness.