Protein language models meet reduced amino acid alphabets.

MOTIVATION: Protein language models (PLMs), which borrowed ideas for modelling and inference from natural language processing, have demonstrated the ability to extract meaningful representations in an unsupervised way. This led to significant performance improvement in several downstream tasks. Clustering amino acids based on their physical-chemical properties to achieve reduced alphabets has been of interest in past research, but their application to PLMs or folding models is unexplored. RESULTS: Here, we investigate the efficacy of PLMs trained on reduced amino acid alphabets in capturing evolutionary information, and we explore how the loss of protein sequence information impacts learned representations and downstream task performance. Our empirical work shows that PLMs trained on the full alphabet and a large number of sequences capture fine details that are lost in alphabet reduction methods. We further show the ability of a structure prediction model(ESMFold) to fold CASP14 protein sequences translated using a reduced alphabet. For 10 proteins out of the 50 targets, reduced alphabets improve structural predictions with LDDT-Cα differences of up to 19%. AVAILABILITY AND IMPLEMENTATION: Trained models and code are available at github.com/Ieremie/reduced-alph-PLM.

Identification of MYCN non-amplified neuroblastoma subgroups points towards molecular signatures for precision prognosis and therapy stratification.

BACKGROUND: Despite the extensive study of MYCN-amplified neuroblastomas, there is a significant unmet clinical need in MYCN non-amplified cases. In particular, the extent of heterogeneity within the MYCN non-amplified population is unknown. METHODS: A total of 1566 samples from 16 datasets were identified in Gene Expression Omnibus (GEO) and ArrayExpress. Characterisation of the subtypes was analysed by ConsensusClusterPlus. Independent predictors for subgrouping were constructed from the single sample predictor based on the multiclassPairs package. Findings were verified using immunohistochemistry and CIBERSORTx analysis. RESULTS: We demonstrate that MYCN non-amplified neuroblastomas are heterogeneous and can be classified into 3 subgroups based on their transcriptional signatures. Within these groups, subgroup_2 has the worst prognosis and this group shows a 'MYCN' signature that is potentially induced by the overexpression of Aurora Kinase A (AURKA); whilst subgroup_3 is characterised by an 'inflamed' gene signature. The clinical implications of this subtype classification are significant, as each subtype demonstrates a unique prognosis and vulnerability to investigational therapies. A total of 420 genes were identified as independent subgroup predictors with average balanced accuracy of 0.93 and 0.84 for train and test datasets, respectively. CONCLUSION: We propose that transcriptional subtyping may enhance precision prognosis and therapy stratification for patients with MYCN non-amplified neuroblastomas.

TransformerGO: predicting protein-protein interactions by modelling the attention between sets of gene ontology terms.

MOTIVATION: Protein-protein interactions (PPIs) play a key role in diverse biological processes but only a small subset of the interactions has been experimentally identified. Additionally, high-throughput experimental techniques that detect PPIs are known to suffer various limitations, such as exaggerated false positives and negatives rates. The semantic similarity derived from the Gene Ontology (GO) annotation is regarded as one of the most powerful indicators for protein interactions. However, while computational approaches for prediction of PPIs have gained popularity in recent years, most methods fail to capture the specificity of GO terms. RESULTS: We propose TransformerGO, a model that is capable of capturing the semantic similarity between GO sets dynamically using an attention mechanism. We generate dense graph embeddings for GO terms using an algorithmic framework for learning continuous representations of nodes in networks called node2vec. TransformerGO learns deep semantic relations between annotated terms and can distinguish between negative and positive interactions with high accuracy. TransformerGO outperforms classic semantic similarity measures on gold standard PPI datasets and state-of-the-art machine-learning-based approaches on large datasets from Saccharomyces cerevisiae and Homo sapiens. We show how the neural attention mechanism embedded in the transformer architecture detects relevant functional terms when predicting interactions. AVAILABILITY AND IMPLEMENTATION: https://github.com/Ieremie/TransformerGO. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Impact of non-pharmaceutical interventions during COVID-19 on future influenza trends in Mainland China.

BACKGROUND: Influenza is a common illness for its high rates of morbidity and transmission. The implementation of non-pharmaceutical interventions (NPIs) during the COVID-19 pandemic to manage its dissemination could affect the transmission of influenza. METHODS: A retrospective analysis, between 2018 and 2023, was conducted to examine the incidence of influenza virus types A and B among patients in sentinel cities located in North or South China as well as in Wuhan City. For validations, data on the total count of influenza patients from 2018 to 2023 were collected at the Central Hospital of Wuhan, which is not included in the sentinel hospital network. Time series methods were utilized to examine seasonal patterns and to forecast future influenza trends. RESULTS: Northern and southern cities in China had earlier outbreaks during the NPIs period by about 8 weeks compared to the 2018-2019. The implementation of NPIs significantly reduced the influenza-like illness (ILI) rate and infection durations. Influenza B Victoria and H3N2 were the first circulating strains detected after the relaxation of NPIs, followed by H1N1 across mainland China. The SARIMA model predicted synchronized H1N1 outbreak cycles in North and South China, with H3N2 expected to occur in the summer in southern cities and in the winter in northern cities over the next 3 years. The ILI burden is expected to rise in both North and South China over the next 3 years, with higher ILI% levels in southern cities throughout the year, especially in winter, and in northern cities mainly during winter. In Wuhan City and the Central Hospital of Wuhan, influenza levels are projected to peak in the winter of 2024, with 2 smaller peaks expected during the summer of 2023. CONCLUSIONS: In this study, we report the impact of NPIs on future influenza trends in mainland China. We recommend that local governments encourage vaccination during the transition period between summer and winter to mitigate economic losses and mortality associated with influenza.

Alternate wiring of a KNOXI genetic network underlies differences in leaf development of A. thaliana and C. hirsuta.

Two interrelated problems in biology are understanding the regulatory logic and predictability of morphological evolution. Here, we studied these problems by comparing Arabidopsis thaliana, which has simple leaves, and its relative, Cardamine hirsuta, which has dissected leaves comprising leaflets. By transferring genes between the two species, we provide evidence for an inverse relationship between the pleiotropy of SHOOTMERISTEMLESS (STM) and BREVIPEDICELLUS (BP) homeobox genes and their ability to modify leaf form. We further show that cis-regulatory divergence of BP results in two alternative configurations of the genetic networks controlling leaf development. In C. hirsuta, ChBP is repressed by the microRNA164A (MIR164A)/ChCUP-SHAPED COTYLEDON (ChCUC) module and ChASYMMETRIC LEAVES1 (ChAS1), thus creating cross-talk between MIR164A/CUC and AS1 that does not occur in A. thaliana. These different genetic architectures lead to divergent interactions of network components and growth regulation in each species. We suggest that certain regulatory genes with low pleiotropy are predisposed to readily integrate into or disengage from conserved genetic networks influencing organ geometry, thus rapidly altering their properties and contributing to morphological divergence.

Bidirectional epithelial-mesenchymal crosstalk provides self-sustaining profibrotic signals in pulmonary fibrosis.

Idiopathic pulmonary fibrosis (IPF) is the prototypic progressive fibrotic lung disease with a median survival of 2 to 4 years. Injury to and/or dysfunction of the alveolar epithelium is strongly implicated in IPF disease initiation, but the factors that determine whether fibrosis progresses rather than normal tissue repair occurs remain poorly understood. We previously demonstrated that zinc finger E-box-binding homeobox 1-mediated epithelial-mesenchymal transition in human alveolar epithelial type II (ATII) cells augments transforming growth factor-ß-induced profibrogenic responses in underlying lung fibroblasts via paracrine signaling. Here, we investigated bidirectional epithelial-mesenchymal crosstalk and its potential to drive fibrosis progression. RNA-Seq of lung fibroblasts exposed to conditioned media from ATII cells undergoing RAS-induced epithelial-mesenchymal transition identified many differentially expressed genes including those involved in cell migration and extracellular matrix regulation. We confirmed that paracrine signaling between RAS-activated ATII cells and fibroblasts augmented fibroblast recruitment and demonstrated that this involved a zinc finger E-box-binding homeobox 1-tissue plasminogen activator axis. In a reciprocal fashion, paracrine signaling from transforming growth factor-ß-activated lung fibroblasts or IPF fibroblasts induced RAS activation in ATII cells, at least partially through the secreted protein acidic and rich in cysteine, which may signal via the epithelial growth factor receptor via epithelial growth factor-like repeats. Together, these data identify that aberrant bidirectional epithelial-mesenchymal crosstalk in IPF drives a chronic feedback loop that maintains a wound-healing phenotype and provides self-sustaining profibrotic signals.

Deconvolution of RNA-Seq Analysis of Hyperbaric Oxygen-Treated Mice Lungs Reveals Mesenchymal Cell Subtype Changes.

Hyperbaric oxygen (HBO) is widely applied to treat several hypoxia-related diseases. Previous studies have focused on the immediate effect of HBO-exposure induced oxidative stress on the lungs, but knowledge regarding the chronic effects from repetitive HBO exposure is limited, especially at the gene expression level. We found that repetitive HBO exposure did not alter the morphology of murine lungs. However, by deconvolution of RNA-seq from those mice lungs using CIBERSORTx and the expression profile matrices of 8 mesenchymal cell subtypes obtained from bleomycin-treated mouse lungs, we identify several mesenchymal cell subtype changes. These include increases in Col13a1 matrix fibroblasts, mesenchymal progenitors and mesothelial cell populations and decreases in lipofibroblasts, endothelial and Pdgfrb high cell populations. Our data suggest that repetitive HBO exposure may affect biological processes in the lungs such as response to wounding, extracellular matrix, vasculature development and immune response.

Proteomic Analysis of Azacitidine-Induced Degradation Profiles Identifies Multiple Chromatin and Epigenetic Regulators Including Uhrf1 and Dnmt1 as Sensitive to Azacitidine.

DNA methylation is a critical epigenetic modification that is established and maintained across the genome by DNA methyltransferase enzymes (Dnmts). Altered patterns of DNA methylation are a frequent occurrence in many tumor genomes, and inhibitors of Dnmts have become important epigenetic drugs. Azacitidine is a cytidine analog that is incorporated into DNA and induces the specific inhibition and proteasomal-mediated degradation of Dnmts. The downstream effects of azacitidine on CpG methylation and on gene transcription have been widely studied in many systems, but how azacitidine impacts the proteome is not well-understood. In addition, with its specific ability to induce the rapid degradation of Dnmts (in particular, the primary maintenance DNA methyltransferase, Dnmt1), it may be employed as a specific chemical knockdown for investigating the Dnmt1-associated functional or physical interactome. In this study, we use quantitative proteomics to analyze the degradation profile of proteins in the nuclear proteome of cells treated with azacitidine. We identify specific proteins as well as multiple pathways and processes that are impacted by azacitidine. The Dnmt1 interaction partner, Uhrf1, exhibits significant azacitidine-induced degradation, and this azacitidine-induced degradation is independent of the levels of Dnmt1 protein. We identify multiple other chromatin- and epigenetic-associated factors, including the bromodomain-containing transcriptional regulator, Brd2. We show that azacitidine induces highly specific perturbations of the Dnmt1-associated proteome, and while interaction partners such as Uhrf1 are sensitive to azacitidine, others such as the Dnmt1 interaction partner and stability regulator, Usp7, are not. In summary, we have conducted the first comprehensive proteomic analysis of the azacitidine-sensitive nuclear proteome, and we show how 5-azacitidine can be used as a specific probe to explore Dnmt- and chromatin-related protein networks.

Multiproteomic and Transcriptomic Analysis of Oncogenic ß-Catenin Molecular Networks.

The dysregulation of Wnt signaling is a frequent occurrence in many different cancers. Oncogenic mutations of CTNNB1/ß-catenin, the key nuclear effector of canonical Wnt signaling, lead to the accumulation and stabilization of ß-catenin protein with diverse effects in cancer cells. Although the transcriptional response to Wnt/ß-catenin signaling activation has been widely studied, an integrated understanding of the effects of oncogenic ß-catenin on molecular networks is lacking. We used affinity-purification mass spectrometry (AP-MS), label-free liquid chromatography-tandem mass spectrometry, and RNA-Seq to compare protein-protein interactions, protein expression, and gene expression in colorectal cancer cells expressing mutant (oncogenic) or wild-type ß-catenin. We generate an integrated molecular network and use it to identify novel protein modules that are associated with mutant or wild-type ß-catenin. We identify a DNA methyltransferase I associated subnetwork that is enriched in cells with mutant ß-catenin and a subnetwork enriched in wild-type cells associated with the CDKN2A tumor suppressor, linking these processes to the transformation of colorectal cancer cells through oncogenic ß-catenin signaling. In summary, multiomics analysis of a defined colorectal cancer cell model provides a significantly more comprehensive identification of functional molecular networks associated with oncogenic ß-catenin signaling.

Proteomics and Network Analyses Reveal Inhibition of Akt-mTOR Signaling in CD4+ T Cells by Mycobacterium tuberculosis Mannose-Capped Lipoarabinomannan.

Mycobacterium tuberculosis (Mtb) cell wall glycolipid mannose-capped lipoarabinomannan (ManLAM) inhibits CD4+ T-cell activation by inhibiting proximal T-cell receptor (TCR) signaling when activated by anti-CD3. To understand the impact of ManLAM on CD4+ T-cell function when both the TCR-CD3 complex and major costimulator CD28 are engaged, we performed label-free quantitative MS and network analysis. Mixed-effect model analysis of peptide intensity identified 149 unique peptides representing 131 proteins that were differentially regulated by ManLAM in anti-CD3- and anti-CD28-activated CD4+ T cells. Crosstalker, a novel network analysis tool identified dysregulated translation, TCA cycle, and RNA metabolism network modules. PCNA, Akt, mTOR, and UBC were found to be bridge node proteins connecting these modules of dysregulated proteins. Altered PCNA expression and cell cycle analysis showed arrest at the G2M phase. Western blot confirmed that ManLAM inhibited Akt and mTOR phosphorylation, and decreased expression of deubiquitinating enzymes Usp9x and Otub1. Decreased NF-κB phosphorylation suggested interference with CD28 signaling through inhibition of the Usp9x-Akt-mTOR pathway. Thus, ManLAM induced global changes in the CD4+ T-cell proteome by affecting Akt-mTOR signaling, resulting in broad functional impairment of CD4+ T-cell activation beyond inhibition of proximal TCR-CD3 signaling.

The CRAPome: a contaminant repository for affinity purification-mass spectrometry data.

Affinity purification coupled with mass spectrometry (AP-MS) is a widely used approach for the identification of protein-protein interactions. However, for any given protein of interest, determining which of the identified polypeptides represent bona fide interactors versus those that are background contaminants (for example, proteins that interact with the solid-phase support, affinity reagent or epitope tag) is a challenging task. The standard approach is to identify nonspecific interactions using one or more negative-control purifications, but many small-scale AP-MS studies do not capture a complete, accurate background protein set when available controls are limited. Fortunately, negative controls are largely bait independent. Hence, aggregating negative controls from multiple AP-MS studies can increase coverage and improve the characterization of background associated with a given experimental protocol. Here we present the contaminant repository for affinity purification (the CRAPome) and describe its use for scoring protein-protein interactions. The repository (currently available for Homo sapiens and Saccharomyces cerevisiae) and computational tools are freely accessible at http://www.crapome.org/.

Proteomic and bioinformatics profile of paired human alveolar macrophages and peripheral blood monocytes.

Little is known about proteomic differences between pluripotent human peripheral blood monocytes (MN) and their terminally-differentiated pulmonary counterparts, alveolar macrophages (AM). To better characterize these cell populations, we performed a label-free shotgun proteomics assessment of matched AM and MN preparations from eight healthy volunteers. With an FDR of less than 0.45%, we identified 1754 proteins within AM and 1445 from MN. Comparison of the two proteomes revealed that 1239 of the proteins found in AM were shared with MN, whereas 206 proteins were uniquely identified in MN and 515 were unique to AM. Molecular and cellular functions, protein classes, development associations, and membership in physiological systems and canonical pathways were identified among the detected proteins. Analysis of biologic processes represented by these proteomes indicated that MN were most prominently enriched for proteins involved in cellular movement and immune cell trafficking. In contrast, AM were enriched for proteins involved in protein trafficking, molecular transport, and cellular assembly and organization. These findings provide a baseline proteomic resource for further studies aimed at better understanding of the functional differences between MN and AM in both health and disease.

How do oncoprotein mutations rewire protein-protein interaction networks?

The acquisition of mutations that activate oncogenes or inactivate tumor suppressors is a primary feature of most cancers. Mutations that directly alter protein sequence and structure drive the development of tumors through aberrant expression and modification of proteins, in many cases directly impacting components of signal transduction pathways and cellular architecture. Cancer-associated mutations may have direct or indirect effects on proteins and their interactions and while the effects of mutations on signaling pathways have been widely studied, how mutations alter underlying protein-protein interaction networks is much less well understood. Systematic mapping of oncoprotein protein interactions using proteomics techniques as well as computational network analyses is revealing how oncoprotein mutations perturb protein-protein interaction networks and drive the cancer phenotype.

DNA and chromatin modification networks distinguish stem cell pluripotent ground states.

Pluripotent stem cells are capable of differentiating into all cell types of the body and therefore hold tremendous promise for regenerative medicine. Despite their widespread use in laboratories across the world, a detailed understanding of the molecular mechanisms that regulate the pluripotent state is currently lacking. Mouse embryonic (mESC) and epiblast (mEpiSC) stem cells are two closely related classes of pluripotent stem cells, derived from distinct embryonic tissues. Although both mESC and mEpiSC are pluripotent, these cell types show important differences in their properties suggesting distinct pluripotent ground states. To understand the molecular basis of pluripotency, we analyzed the nuclear proteomes of mESCs and mEpiSCs to identify protein networks that regulate their respective pluripotent states. Our study used label-free LC-MS/MS to identify and quantify 1597 proteins in embryonic and epiblast stem cell nuclei. Immunoblotting of a selected protein subset was used to confirm that key components of chromatin regulatory networks are differentially expressed in mESCs and mEpiSCs. Specifically, we identify differential expression of DNA methylation, ATP-dependent chromatin remodeling and nucleosome remodeling networks in mESC and mEpiSC nuclei. This study is the first comparative study of protein networks in cells representing the two distinct, pluripotent states, and points to the importance of DNA and chromatin modification processes in regulating pluripotency. In addition, by integrating our data with existing pluripotency networks, we provide detailed maps of protein networks that regulate pluripotency that will further both the fundamental understanding of pluripotency as well as efforts to reliably control the differentiation of these cells into functional cell fates.

Human biomarker discovery and predictive models for disease progression for idiopathic pneumonia syndrome following allogeneic stem cell transplantation.

Allogeneic hematopoietic stem cell transplantation (SCT) is the only curative therapy for many malignant and nonmalignant conditions. Idiopathic pneumonia syndrome (IPS) is a frequently fatal complication that limits successful outcomes. Preclinical models suggest that IPS represents an immune mediated attack on the lung involving elements of both the adaptive and the innate immune system. However, the etiology of IPS in humans is less well understood. To explore the disease pathway and uncover potential biomarkers of disease, we performed two separate label-free, proteomics experiments defining the plasma protein profiles of allogeneic SCT patients with IPS. Samples obtained from SCT recipients without complications served as controls. The initial discovery study, intended to explore the disease pathway in humans, identified a set of 81 IPS-associated proteins. These data revealed similarities between the known IPS pathways in mice and the condition in humans, in particular in the acute phase response. In addition, pattern recognition pathways were judged to be significant as a function of development of IPS, and from this pathway we chose the lipopolysaccaharide-binding protein (LBP) protein as a candidate molecular diagnostic for IPS, and verified its increase as a function of disease using an ELISA assay. In a separately designed study, we identified protein-based classifiers that could predict, at day 0 of SCT, patients who: 1) progress to IPS and 2) respond to cytokine neutralization therapy. Using cross-validation strategies, we built highly predictive classifier models of both disease progression and therapeutic response. In sum, data generated in this report confirm previous clinical and experimental findings, provide new insights into the pathophysiology of IPS, identify potential molecular classifiers of the condition, and uncover a set of markers potentially of interest for patient stratification as a basis for individualized therapy.

LKB1 depletion-mediated epithelial-mesenchymal transition induces fibroblast activation in lung fibrosis.

The factors that determine fibrosis progression or normal tissue repair are largely unknown. We previously demonstrated that autophagy inhibition-mediated epithelial-mesenchymal transition (EMT) in human alveolar epithelial type II (ATII) cells augments local myofibroblast differentiation in pulmonary fibrosis by paracrine signalling. Here, we report that liver kinase B1 (LKB1) inactivation in ATII cells inhibits autophagy and induces EMT as a consequence. In IPF lungs, this is caused by downregulation of CAB39L, a key subunit within the LKB1 complex. 3D co-cultures of ATII cells and MRC5 lung fibroblasts coupled with RNA sequencing (RNA-seq) confirmed that paracrine signalling between LKB1-depleted ATII cells and fibroblasts augmented myofibroblast differentiation. Together these data suggest that reduced autophagy caused by LKB1 inhibition can induce EMT in ATII cells and contribute to fibrosis via aberrant epithelial-fibroblast crosstalk.

Synthetic lethal approaches to target cancers with loss of PTEN function.

Phosphatase and tensin homolog (PTEN) is a tumour suppressor gene and has a role in inhibiting the oncogenic AKT signalling pathway by dephosphorylating phosphatidylinositol 3,4,5-triphosphate (PIP3) into phosphatidylinositol 4,5-bisphosphate (PIP2). The function of PTEN is regulated by different mechanisms and inactive PTEN results in aggressive tumour phenotype and tumorigenesis. Identifying targeted therapies for inactive tumour suppressor genes such as PTEN has been challenging as it is difficult to restore the tumour suppressor functions. Therefore, focusing on the downstream signalling pathways to discover a targeted therapy for inactive tumour suppressor genes has highlighted the importance of synthetic lethality studies. This review focuses on the potential synthetic lethality genes discovered in PTEN-inactive cancer types. These discovered genes could be potential targeted therapies for PTEN-inactive cancer types and may improve the treatment response rates for aggressive types of cancer.

Integrated network pharmacology and cellular assay reveal the biological mechanisms of Limonium sinense (Girard) Kuntze against Breast cancer.

BACKGROUND: Limonium Sinense (Girard) Kuntze (L. sinense) has been widely used for the treatment of anaemia, bleeding, cancer, and other disorders in Chinese folk medicine. The aim of this study is to predict the therapeutic effects of L. sinense and investigate the potential mechanisms using integrated network pharmacology methods and in vitro cellular experiments. METHODS: The active ingredients of L. sinense were collected from published literature, and the potential targets related to L. sinense were obtained from public databases. Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) and DisGeNET enrichment analyses were performed to explore the underlying mechanisms. Molecular docking, cellular experiments, RNA-sequencing (RNA-seq) and Gene Expression Omnibus (GEO) datasets were employed to further evaluate the findings. RESULTS: A total of 15 active ingredients of L. sinense and their corresponding 389 targets were obtained. KEGG enrichment analysis revealed that the biological effects of L. sinense were primarily associated with "Pathways in cancer". DisGeNET enrichment analysis highlighted the potential role of L. sinense in the treatment of breast cancer. Apigenin within L. sinense showed promising potential against cancer. Cellular experiments demonstrated that the L. sinense ethanol extract (LSE) exhibited a significant growth inhibitory effect on multiple breast cancer cell lines in both 2D and 3D cultures. RNA-seq analysis revealed a potential impact of LSE on breast cancer. Additionally, analysis of GEO datasets verified the significant enrichment of breast cancer and several cancer-related pathways upon treatment with Apigenin in human breast cancer cells. CONCLUSION: This study predicts the biological activities of L. sinense and demonstrates the inhibitory effect of LSE on breast cancer cells, highlighting the potential application of L. sinense in cancer treatment.