Biology subjects

Zhang, S.-W.

Publications and source records attributed to Zhang, S.-W..

7 recordsLinked to original sources

bioRxiv2026

Causally-inspired meta-representation learning framework for predicting patient-specific clinical responses to drug combinations

Large-scale prediction and assessment of clinical patient responses (i.e., RECIST class) to drug combinations remains challenging due to scarce patient-derived data. The existing prediction methods mainly rely on cancer cell line models. However, substantial biological heterogeneity between cancer cell lines and cancer patients within same tissues, as well as the heterogeneity between one tissue and another, often limit the generalizability of these methods in clinical patients. To overcome these limitations, here we present CaMeRe, a Causally-inspired Meta-representation learning framework designed to predict patient-specific clinical Response to drug combinations. In situations where stable causal factors and domain-specific response-modulating factors are unobservable, explicit discrete domain labels are unavailable, and data is scarce, CaMeRe designed a domain-invariant causal representation learning (DICRL) model guided by the invariant information bottleneck theory and causal intervention invariance principle, and also built a meta-learning framework with bi-level domain generalization to optimize DICRL model for achieving multi-domain generalization within and across tissues. By integrating the causal representation learning and meta learning framework, CaMeRe not only exhibited robust multi-domain generalization performance across multiple clinical drug combination response datasets and PDXs drug combination response datasets and generalization scenarios, but also had better interpretability. We applied CaMeRe to predict drug-combination response scores for 3,423 patients across 542,080 drug combinations. The predicted scores were significantly associated with biomarkers of known drug combinations and enabled the prioritization of candidate drug combinations across 11 cancer types, with stronger support from literature and clinical trial evidences than random baselines. We believe that CaMeRe can be a useful tool for predicting large-scale clinical individual drug combination responses and it has broad clinical applications.

bioinformatics↗

bioRxiv2024

Uncovering topologically associating domains from three-dimensional genome maps with TADGATE

Topologically associating domains (TADs) emerge as indispensable units in three-dimensional (3D) genome organization, playing a critical role in gene regulation. However, accurately identifying TADs from sparse chromatin contact maps and exploring the structural and functional elements within TADs remain challenging. To this end, we develop a graph attention auto-encoder, TADGATE, to accurately identify TADs even from ultra-sparse contact maps and generate the imputed maps while preserving or enhancing the underlying topological structures. TADGATE can capture specific attention patterns, pointing to two types of units with different characteristics in TADs. Moreover, we find that the organization of TADs is closely associated with chromatin compartmentalization, and TAD boundaries in different compartmental environments exhibit distinct biological properties. We also utilize a two-layer Hidden Markov Model to functionally annotate the TADs and their internal regions, revealing the overall properties of TADs and the distribution of the structural and functional elements within TADs. At last, we apply TADGATE to highly sparse and noisy Hi-C contact maps from 21 human tissues or cell lines, enhancing the clarity of TAD structures, investigating the nature of conserved and cell type-specific boundaries, and unveiling the cell type-specific transcriptional regulatory mechanisms associated with topological domains.

bioinformatics↗

bioRxiv2022

Defining the separation landscape of topological domains for decoding consensus domain organization of 3D genome

Topologically associating domains (TADs) have emerged as basic structural and functional units of genome organization, and have been determined by many computational methods from Hi-C contact maps. However, the TADs obtained by different methods vary greatly, which makes the accurate determination of TADs a challenging issue and hinders subsequent biological analyses about their organization and functions. Obvious inconsistencies among the TADs identified by different methods indeed make the statistical and biological properties of TADs overly depend on the method we chose rather than on the data. To this end, we employ the consensus structural information captured by these methods to define the TAD separation landscape for decoding consensus domain organization of the 3D genome. We demonstrate that the TAD separation landscape could be used to compare domain boundaries across multiple cell types for discovering conserved and divergent topological structures, decipher three types of boundary regions with diverse biological features, and identify Consensus Topological Associating Domains (ConsTADs). We illustrate that these analyses could deepen our understanding of the relationships between the topological domains and chromatin states, gene expression, and DNA replication timing. In short, we provide an alternative solution to deal with the serious inconsistencies of TADs obtained via different methods by defining the TAD separation landscape and ConsTAD.

bioinformatics↗

bioRxiv2022

DGMP: Identifying Cancer Driver Genes by Jointing DGCN and MLP from Multi-omics Genomic Data

Identification of cancer driver genes plays an important role in precision oncology research, which is helpful to understand the cancer initiation and progression. However, most of existing computational methods mainly used the protein-protein interaction networks (PPIs), or treated the directed gene regulatory networks (GRNs) as the undirected gene-gene association networks to identify the cancer driver genes, which will lose the unique structure regulatory information in the directed GRNs, and then affect the outcome of the cancer driver genes identification. Here, based on the multi-omics pan-cancer data (i.e., gene expression, mutation, copy number variation and DNA methylation), we proposed a novel method (called DGMP) to identify cancer driver genes by jointing Directed Graph Convolution Network (DGCN) and Multilayer Perceptron (MLP). DGMP learns the multi-omics features of genes as well as the topological structure features in GRN with DGCN model, and uses MLP to weight more on gene features for mitigating the bias toward the graph topological features in DGCN learning process. The results on three gene regulation networks show that DGMP outperforms other existing state-of-the-art methods. It can not only identify highly mutated cancer driver genes but also the driver genes harboring other kinds of alterations (e.g., differential expression, aberrant DNA methylation) or genes involved in GRNs with other cancer genes. The source code of DGMP can be freely downloaded from https://github.com/NWPU-903PR/DGMP.

bioinformatics↗

bioRxiv2021

Arabidopsis ROOT UV-B SENSITIVE 1 and 2 Interact with Aminotransferases to Regulate Vitamin B6 Homeostasis

Pyridoxal-5-phosphate (PLP), the enzymatic cofactor form of Vitamin B6 (vitB6), is a versatile compound that has essential roles in metabolism. Cellular PLP homeostasic regulation is currently not well understood. Here we report that in Arabidopsis, biosynthesized PLP is sequestered by specific aminotransferases (ATs), and that the proteins ROOT UV-B SENSITIVE 1 (RUS1) and RUS2 function with ATs to regulate PLP homeostasis. The stunted growth phenotypes of rus1 and rus2 mutants were previously shown to be rescuable by exogenously supplied vitB6. Specific residue changes near the PLP-binding pocket in ASPARTATE AMINOTRANSFERASE2 (ASP2) also rescued rus1 and rus2 phenotypes. In this study, saturated suppressor screens identified 14 additional suppressor of rus (sor) alleles in four aminotransferase genes (ASP1, ASP2, ASP3, or ALANIN AMINOTRANSFERASE1 (AAT1)), which suppressed the rus phenotypes to varying degrees. Each of the sor mutations altered an amino acid in the PLP-binding pocket of the protein, and sor proteins were found to have reduced levels of PLP conjugation. Genetic data revealed that the availability of PLP normally requires both RUS1 and RUS2, and that increasing the number of sor mutants additively enhanced the suppression of rus phenotypes. Biochemical results showed that RUS1 and RUS2 physically interacted with ATs. Our studies suggest a mechanism in which RUS1, RUS2 and specific ATs work together to regulate PLP homeostasis in Arabidopsis.

genetics↗

bioRxiv2019

Identifying drug combinations of individual cancer patients by personalized drug controller method

Treating cancer in precision medicine, it is important to identify the personalized combinational drugs under consideration of the individual heterogeneity. Many bioinformatics tools for the personalized driver genes identification have presented promising clues in determining candidate personalized drug targets for the personalized drugs discovery. However, it has not been studied how to fill the gap between personalized driver genes identification and personalized combinational drugs discovery. In this work, we developed a novel algorithm of structure network Controllability-based Personalized driver Genes and combinational Drug identification (CPGD), aiming to mine the personalized driver genes and identify the combinational drugs of an individual cancer patient. On two benchmark cancer datasets, the performance of CPGD for predicting the clinical efficacious combinational drugs is superior to that of other state-of-the-art driver gene-focus algorithms in terms of precision accuracy. In particular, by quantifying and referring the relationships between target genes of pairwise combinatorial drugs and disease module genes on breast cancer data set, CPGD can significantly divide patients into the discriminative high-risk and low-risk groups for risk asessment in combination therapy. In addition, CPGD can further enhance cancer subtyping by providing computationally personalized side effect signatures for individual patients. Collectively, CPGD provided a new and effecient bioinformatics tool from structure network controllability perspective for discovering personalized combinational drugs with personalized side effect consideration, so as to effectively support personalized risk assessement and disease subtyping. SignificanceIt is quite challenging to predict personalized combinational drugs rather than patient-cohort s drugs based on cancer omics data. In this work, a novel structure network Controllability-based algorithm (CPGD) from feedback vertex sets control perspective was developed, for discovering efficacious combinational drugs of an individual cancer patient by targeting the personalized driver genes. The CPGD contains three methodological advances by exploring more precise mathematical models on high-throughput personalized multi-omics data. The first is that a proper network structure is constructed to characterize the gene regulatory mechanism of an individual patient. The second is that considering the weight information of network edges/relations improves the performance for predicting clinical efficacious combinational drugs compared with other drivers-focus methods. And the third is that proper evaluation metrics for personalized combinational drugs prioritization, personalized risk assessment and disease subtyping are designed when evaluating the performance of CPGD.

systems biology↗

bioRxiv2018

A novel structure-based control method for analyzing nonlinear dynamics in biological networks

Exploring complex biological systems requires adequate knowledge of the systems underlying wiring diagram but not its specific functional forms. Thus, exploration actually requires the concepts and approaches delivered by structure-based network control, which investigates the controllability of complex networks through a minimum set of input nodes. Traditional structure-based control methods focus on the structure of complex systems with linear dynamics and may not match the meaning of control well in some biological systems. Here we took into consideration the nonlinear dynamics of some biological networks and formalized the nonlinear control problem of undirected dynamical networks (NCU). Then, we designed and implemented a novel and general graphic-theoretic algorithm (NCUA) from the perspective of the feedback vertex set to discover the possible minimum sets of the input nodes in controlling the network state. We applied our NCUA to both synthetic networks and real-world networks to investigate how the network parameters, such as the scaling exponent and the degree heterogeneity, affect the control characteristics of networks with nonlinear dynamics. The NCUA was applied to analyze the patient-specific molecular networks corresponding to patients across multiple datasets from The Cancer Genome Atlas (TCGA), which demonstrates the advantages of the nonlinear control method to characterize and quantify the patient-state change over the other state-of-the-art linear control methods. Thus, our model opens a new way to control the undesired transition of cancer states and provides a powerful tool for theoretical research on network control, especially in biological fields.\n\nAuthor summaryComplex biological systems usually have nonlinear dynamics, such as the biological gene (protein) interaction network and gene co-expression networks. However, most of the structure-based network control methods focus on the structure of complex systems with linear dynamics. Thus, the ultimate purpose to control biological networks is still too complicated to be directly solved by such network control methods. We currently lack a framework to control the biological networks with nonlinear and undirected dynamics theoretically and computationally. Here, we discuss the concept of the nonlinear control problem of undirected dynamical networks (NCU) and present the novel graphic-theoretic algorithm from the perspective of a feedback vertex set for identifying the possible sets with minimum input nodes in controlling the networks. The NCUA searches the minimum set of input nodes to drive the network from the undesired attractor to the desired attractor, which is different from conventional linear network control, such as that found in the Maximum Matching Sets (MMS) and Minimum Dominating Sets (MDS) algorithms. In this work, we evaluated the NCUA on multiple synthetic scale-free networks and real complex networks with nonlinear dynamics and found the novel control characteristics of the undirected scale-free networks. We used the NCUA to thoroughly investigate the sample-specific networks and their nonlinear controllability corresponding to cancer samples from TCGA which are enriched with known driver genes and known drug target as controls of pathologic phenotype transitions. We found that our NCUA control method has a better predicted performance for indicating and quantifying the patient biological system changes than that of the state-of-the-art linear control methods. Our approach provides a powerful tool for theoretical research on network control, especially in a range of biological fields.

systems biology↗