Biology subjects

Yi, S. S.

Publications and source records attributed to Yi, S. S..

14 recordsLinked to original sources

bioRxiv2026

Unveiling the Hidden Rules: Enhancing NMD Prediction for Protein-Truncating Variants

Nonsense-mediated decay (NMD) is a conserved RNA quality-control pathway that degrades transcripts containing premature termination codons. Because roughly a third of pathogenic variants in ClinVar can lead to truncated protein synthesis, predicting whether such transcripts undergo NMD is central to interpreting variant effects, yet the canonical 50-55 nucleotide rule explains only about half of observed outcome variability. Using paired whole-genome and RNA-sequencing from 10,306 individual samples in the Trans-Omics for Precision Medicine (TOPMed) program, we quantified NMD efficiency for 5,749 germline truncating variants via allele-specific expression and trained a gradient-boosting classifier, TrunCat, that distinguished NMD-sensitive from NMD-escape transcripts with [~]78% ROC-AUC (Receiver Operating Characteristic - Area Under the Curve). A reduced model using the ten features with the highest mean SHAP (SHapley Additive exPlanations) value as a measure of each features average contribution to predictions nearly matched this performance. Applied across large variant databases and a rare-disease cohort, the model produced NMD outcome predictions, with variants of uncertain significance showing higher predicted escape than pathogenic ones. This framework confirms the canonical rule, identifies non-canonical determinants, and offers a scalable resource for interpreting protein-truncating variants.

genomics↗

bioRxiv2026

eSIG-Net: Accurate prediction of single-mutation induced perturbations on protein interactions using a language model

Most proteins exert their functions in complex with other interactors. Single mutations can exhibit a profound impact on perturbing protein interactions, leading to human disease. However, predicting the effect of single mutations on protein interactions remains a major computational challenge. Deep learning, particularly protein language models or transformers, has become an effective tool in bioinformatics for protein structure prediction. However, the functional divergence of mutations makes it difficult to predict their interaction perturbation profiles. To address this fundamental challenge, we present eSIG-Net (edgetic mutation Sequence-based Interaction Grammar Network), a novel sequence-based "Interaction Language Model" for predicting protein interaction alterations caused by single mutations. eSIG-Net combines various protein sequence embeddings, introduces a mutation-encoding module with syntax and evolutionary insights, and employs contrastive learning to evaluate mutation-induced interaction changes. eSIG-Net significantly outperforms current state-of-the-art sequence-based and structure-based prediction methods at predicting mutational impact on protein interactions. We highlight examples where eSIG-Net nominates causal variants with high confidence and elucidates their functional role under relevant biological contexts. Together, eSIG-Net is a first-in-kind "interaction language model" that can accurately predict interaction-specific rewiring by single mutations with only sequence information, and exhibits generalizability across biological contexts.

bioinformatics↗

bioRxiv2025

Viral transcriptional regulators extensively rewire host pathways through diverse mechanisms

Viral transcriptional regulators (vTRs) reprogram host gene regulatory networks to promote replication, persistence, and immune evasion. Despite the identification of hundreds of vTRs in human viruses, how they rewire host pathways remains unclear. Here, we systematically profiled 95 vTRs from diverse human viruses across multiple functional assays. vTRs perturb immune, cell proliferation/death, and signaling pathways through various mechanisms; some bind DNA directly, others cooperate or antagonize human transcription factors (hTFs), and some remodel chromatin. vTRs can act as activators or repressors and recruit similar but not identical repertoires of proteins as hTFs. These findings reveal vTRs as versatile transcriptional modulators that converge on conserved host "pressure points" while diversifying across pathways to promote viral replication and persistence. Notably, many vTR dysregulate genes within autoimmune, neurological, and cardiovascular risk loci, revealing mechanistic links to disease. Together, we provide a comprehensive resource for understanding and targeting viral control of human transcription.

genomics↗

bioRxiv2025

Nuclear Phase Separation Drives NPM1-mutant Acute Myeloid Leukemia

During cancer development, mutations promote gene expression changes that cause transformation. Leukemia is frequently associated with aberrant HOXA expression driven by translocations in nucleoporin genes or KMT2A, and mutations in NPM1. How disparate mutations converge on this regulatory pathway is not understood. Here we demonstrate that mutant NPM1 (NPM1c) forms nuclear condensates in multiple human cell lines, mouse models, and primary patient samples. We show NPM1c phase separation is necessary and sufficient to coordinate the recruitment of NUP98 and KMT2A to condensates. Through extensive mutagenesis and pharmacological destabilization of phase separation, we find that NPM1c condensates are necessary for regulating gene expression, promoting in vivo expansion, and maintaining the undifferentiated leukemic state. Finally, we show that nucleoporin and KMT2A fusion proteins form condensates that are biophysically indistinguishable from NPM1c condensates. Together, these data define a new condensate underlying leukemias that we term coordinating bodies (C-bodies), and propose C-bodies as a therapeutic vulnerability.

biophysics↗

bioRxiv2025

Therapeutic Potential of PRMT1 as a Critical Survival Dependency Target in Multiple Myeloma

Multiple myeloma (MM) is a neoplasm of antibody-producing plasma cells and is the second most prevalent hematological malignancy worldwide. Development of drug resistance and disease relapse significantly impede the success of MM treatment, highlighting the critical need to discover novel therapeutic targets. In a custom CRISPR/Cas9 screen targeting 197 DNA damage response-related genes, Protein Arginine N-Methyltransferase 1 (PRMT1) emerged as a top hit, revealing it as a potential therapeutic vulnerability and survival dependency in MM cells. PRMT1, a major Type I PRMT enzyme, catalyzes the asymmetric transfer of methyl groups to arginine residues, influencing gene transcription and protein function through post-translational modification. Dysregulation or overexpression of PRMT1 has been observed in various malignancies including MM and is linked to chemoresistance. Treatment with the Type I PRMT inhibitor GSK3368715 resulted in a dose-dependent reduction in cell survival across a panel of MM cell lines. This was accompanied by reduced levels of asymmetric dimethylation of arginine (ADMA) and increased arginine monomethylation (MMA) in MM cells. Cell cycle analysis revealed an accumulation of cells in the G0/G1 phase and a reduction in the S phase upon GSK3368715 treatment. Additionally, PRMT1 inhibition led to a significant downregulation of genes involved in cell proliferation, DNA replication, and DNA damage response (DDR), likely inducing genomic instability and impairing tumor growth. This was supported by Reverse Phase Protein Array (RPPA) analyses, which revealed a significant reduction in levels of proteins associated with cell cycle regulation and DDR pathways. Overall, our findings indicate that MM cells critically depend on PRMT1 for survival, highlighting the therapeutic potential of PRMT1 inhibition in treating MM.

cancer biology↗

bioRxiv2023

Population-level comparisons of gene regulatory networks modeled on high-throughput single-cell transcriptomics data.

Single-cell technologies enable high-resolution studies of phenotype-defining molecular mechanisms. However, data sparsity and cellular heterogeneity make modeling biological variability across single-cell samples difficult. We present SCORPION, a tool that uses a message-passing algorithm to reconstruct comparable gene regulatory networks from single cell/nuclei RNA-seq data that are suitable for population-level comparisons by leveraging the same baseline priors. Using synthetic data, we found that SCORPION outperforms 12 other gene regulatory network reconstruction techniques. Using supervised experiments, we show that SCORPION can accurately identify differences in regulatory networks between wild-type and transcription factor-perturbed cells. We demonstrate SCORPIONs scalability to population-level analyses using a single-cell RNA-seq atlas containing 200,436 cells from colorectal cancer and adjacent healthy tissues. The differences detected by SCORPION between tumor regions are consistent across population cohorts, as well as with our understanding of disease progression and elucidate phenotypic regulators that may impact patient survival.

bioinformatics↗

bioRxiv2022

Rapid iPSC inclusionopathy models shed light on formation, consequence and molecular subtype of alpha-synuclein inclusions

Intracellular inclusions accompanying neurodegeneration are histopathologically and ultrastructurally heterogeneous but the significance of this heterogeneity is unclear. iPSC models, while promising for disease modeling, do not form inclusions in a reasonable timeframe and suffer from limited tractability. Here, we developed an iPSC toolbox utilizing piggyBac-based or targeted transgenes to rapidly induce CNS cells with concomitant expression of aggregation-prone proteins. This system is amenable to screening and longitudinal tracking at single-cell and single-inclusion resolution. For proof-of-principle, cortical neuron -synuclein "inclusionopathy" models were engineered to form inclusions through exogenous seeding or -synuclein mutation. These models recapitulated known fibril- and lipid-rich inclusion subtypes, uncovering dynamic interactions between them, and refined the classification of inclusions in postmortem brain. Genetic-modifier and protein-interaction screens pinpointed proteins like RhoA whose sequestration into specific inclusion subtypes is likely to be toxic. This iPSC platform should enhance our understanding of proteinaceous pathologies in neurodegeneration and facilitate therapeutics development.

neuroscience↗

bioRxiv2022

Beyond BRCA: Discovery of novel causes and consequences of homologous recombination deficiencies

Since the discovery of BRCA1 and BRCA2 mutations as cancer risk factors, we have gained substantial insight into their role in maintaining genomic stability through homologous recombination (HR) DNA repair. However, upon pan-cancer analysis of tumors from The Cancer Genome Atlas (TCGA), we found that mutations in BRCA1/2 and other classical HR genes only identified 10-20% of tumors that display genomic evidence of HR deficiency (HRD), suggesting that the cause of the vast majority of HR defects in tumors is unknown. As HRD both predisposes individuals to cancer development and leads to therapeutic vulnerabilities, it is critical to define the spectrum of genetic events that drive HRD. Here, we employed a network-based approach leveraging the abundance of molecular characterization data from TCGA to identify novel drivers of HRD. We discovered that over half of putative genes driving HRD originated outside of canonical DNA damage response genes, with a particular enrichment for RNA binding protein (RBP)-encoding genes. These novel drivers of HRD were cross-validated using an independent ICGC cohort, and were enriched in GWAS loci associated with cancer risk. Experimental approaches validated over 90% of our predictions in a panel of 50 genes tested by siRNA and 31 additional engineered mutations identified from TCGA patient tumors. Moreover, genetic suppression of identified RBPs or pharmacological inhibition of RBPs induced PARP inhibition. Further mechanistic studies indicate that some RBPs are recruited to sites of DNA damage to facilitate repair, whereas others control the expression of canonical HR genes. Overall, this study greatly expands the repertoire of known drivers of HRD and their contributions to DNA damage repair, which has implications for not only future mechanistic studies, but also for genetic screening and therapy stratification. HIGHLIGHTSO_LIThe majority of HR deficiencies detected cannot be directly attributed to aberrations in canonical HR genes. C_LIO_LIIntegrated network analysis identifies RNA binding proteins (RBPs) as a novel driver of HR deficiency in patient tumors. C_LIO_LIRBP dysfunction can produce HR deficiencies through both dysregulation of canonical HR genes and action at sites of DNA damage. C_LI

systems biology↗

bioRxiv2022

Drug combination prioritization for cancer treatment using single-cell RNA-seq based transfer learning

Precision oncology seeks to match patients to the optimal pharmacological regimen; yet, due to tumor heterogeneity, this is challenging. Numerous studies have been conducted to produce clinically relevant pharmacological response forecasts by integrating modern machine learning algorithms and several data types. Insufficient patient numbers and lack of knowledge of the molecular targets for each drug under study limit their use. As a proof of concept, we use single-cell RNA-seq based transfer learning to contextualize patients tumor cells in terms of their more similar cell lines with known susceptibility to drug combinations. Our objective is to maximize the translational potential of in-vitro assays for identifying synergistic drug combinations and prioritizing them for clinical use. Consistent findings in a cohort of breast cancer patients corroborated our understanding of the diseases molecular subtypes. To aid in creating personalized treatments and data-driven clinical trials, we identified the most prevalent cell lines and prioritized synergistic combinations based on tumor compositions at various resolution levels.

bioinformatics↗

bioRxiv2022

Deep learning based on multi-omics integration identifies potential therapeutic targets in breast cancer

Effective and precise classification of breast cancer patients for their disease risks is critical to improve early diagnosis and patient survival. In the recent past, a significant amount of multi-omics data derived from cancer patients has emerged. However, a robust framework for integrating multi-omics data to subgroup cancer patients and predict survival prognosis is still lacking. In addition, effective therapeutic targets for treating breast cancer patients with poor prognoses are in dire need. To begin to resolve this difficulty, we developed and optimized a sophisticated deep learning-based model in breast cancer that can accurately stratify patients based on their prognosis. We built a survival-associated predictive framework integrating transcription profile, miRNA expression, somatic mutations, copy number variation, DNA methylation and protein expression. This framework achieved promising performance in distinguishing high-risk breast cancer patients from those with good prognoses. Furthermore, we constructed multiple fully connected neural networks that are trained on prioritized multi-omics signatures or even only potential single-omics signatures, based on our customized scoring system. Together, the landmark multi-omics signatures we identified may serve as potential therapeutic targets in breast cancer.

bioinformatics↗

bioRxiv2021

Single-cell transcriptomics elucidates in vitro reprogramming of human intestinal epithelium cultured in a physiodynamic gut-on-a-chip

The microphysiological human gut-on-a-chip has demonstrated in vivo-relevant cellular fidelity of intestinal epithelium compared to its cultures in a static condition1, 2. Microfluidic control of morphogen gradients and mechanical cues robustly induced morphological histogenesis with villi-like three-dimensional (3D) microarchitecture, lineage-associated cytodifferentiation, and physiological functions of a human intestinal Caco-2 epithelium3, 4. However, transcriptomic dynamics that orchestrates morphological and functional reprogramming of the epithelium in a microphysiological culture remains elusive. Single-cell transcriptomic analysis revealed that a gut-on-a-chip culture that offers physiological motions and flow drives three distinctive subclusters that offer distinct gene expression and unique spatial representation in 3D epithelial layers. The pseudotemporal trajectory of individual cells visualized the evolutionary transition from ancestral genotypes in static cultures into more heterogeneous phenotypes in physiodynamic cultures on cell cycles, differentiation, and intestinal functions including digestion, absorption, drug transport, and metabolism of xenobiotics. Furthermore, the inversed transcriptomic signature of oncogenes and tumor-suppressor genes of Caco-2 cells verified that a gut-on-a-chip culture drives a postmitotic reprogramming of cancer-associated phenotypes. Thus, we discovered that a physiodynamic on-chip culture is necessary and sufficient for a cancer cell line to be reprogrammed to elicit in vivo-relevant heterogeneous cell populations with restored normal physiological signatures.

bioengineering↗

bioRxiv2020

e-MutPath: Computational modelling reveals the functional landscape of genetic mutations rewiring interactome networks

Understanding the functional impact of cancer somatic mutations represents a critical knowledge gap for implementing precision oncology. It has been increasingly appreciated that the edgotype of a genomic mutation provides a fundamental link between genotype and phenotype. However, specific effects on biological signaling networks for the majority of mutations are largely unknown by experimental approaches. To resolve this challenge, we developed e-MutPath, a network-based computational method to identify candidate edgetic mutations that perturb functional pathways. e-MutPath identifies informative paths that could be used to distinguish disease risk factors from neutral elements and to stratify disease subtypes with clinical relevance. The predicted targets are enriched in cancer vulnerability genes, known drug targets but depleted for proteins associated with side effects, demonstrating the power of network-based strategies to investigate the functional impact and perturbation profiles of genomic mutations. Together, e-MutPath represents a robust computational tool to systematically assign functions to genetic mutations, especially in the context of their specific pathway perturbation effect. The code for e-MutPath is available as a user-friendly R package at the GitHub website (https://github.com/lyshaerbin/eMutPath).

systems biology↗

bioRxiv2020

Comprehensive and Integrated Genomic Characterization of Human Immunome in Cancer

Genetic alterations in immune-related pathways are common hallmarks of cancer. However, to realize the full potential of immunotherapy, a comprehensive understanding of immune networks and how mutations impact network structure and functional output across cancer types is instrumental. Herein we systematically interrogated somatic mutations that could express neoantigens and alter immune responses in cancer patients compared to wild-type controls. To do so, we developed a network-based immunogenomics model (NIPPER) with scoring systems to prioritize critical genes and mutations eliciting differential HLA binding affinity and alternate responses to immunotherapy. These mutations are enriched in essential protein domains and often alter tumor infiltration by immune cells, affecting T cell receptor repertoire and B cell clonal expansion. Furthermore, we devised an interactome network propagation framework integrated with drug associated gene signatures to identify potential immunomodulatory drug candidates. Together, our systems-level analysis results help interpret the heterogeneous immune responses among patients, and serve as a resource for future functional studies and targeted therapeutics. SignificanceCancer cells induce specific immune-related pathway perturbations by mutations, transcriptional dysregulation, and integration of multi-omics data can help identify critical molecular determinants for effective targeted therapeutics.

immunology↗

bioRxiv2020

Systematic characterization of somatic mutation-mediated microRNA regulatory network perturbations

Somatic mutations are a major source of cancer development. Many driver mutations have been identified in protein coding regions. However, the function of mutations located in microRNAs (miRNAs) and their target binding sites along the human genome remains largely unknown. Here, we built comprehensive cancer-specific miRNA regulatory networks across 30 cancer types to systematically analyze the effect of mutations on miRNA related pathways. 3,518,261 mutations from 9,819 samples were mapped to miRNA-gene interactions (mGI), and mutations in miRNAs versus in their target genes show a mutually exclusive pattern in almost all cancer types. Using a linear regression method, we further identified 89 driver mutations in 14 cancer types that can significantly perturb miRNA regulatory networks. We find that driver mutations play their roles by altering RNA binding energy and the expression of target genes. Finally, we demonstrate that mutated driver gene targets are significantly down-regulated in cancer and function as tumor suppressors during cancer progression, suggesting potential miRNA candidates with significant clinical implications. We provide this data resource (CanVar-mGI) through a user-friendly, open-access web portal. Together, our results will facilitate novel non-coding biomarker identification and therapeutic drug design.

bioinformatics↗