[Identification of Protein-Coding Gene Markers in Breast Invasive Carcinoma Based on Machine Learning].

Objective To screen out the biomarkers linked to prognosis of breast invasive carcinoma based on the analysis of transcriptome data by random forest (RF),extreme gradient boosting (XGBoost),light gradient boosting machine (LightGBM),and categorical boosting (CatBoost). Methods We obtained the expression data of breast invasive carcinoma from The Cancer Genome Atlas and employed DESeq2,t-test,and Cox univariate analysis to identify the differentially expressed protein-coding genes associated with survival prognosis in human breast invasive carcinoma samples.Furthermore,RF,XGBoost,LightGBM,and CatBoost models were established to mine the protein-coding gene markers related to the prognosis of breast invasive cancer and the model performance was compared.The expression data of breast cancer from the Gene Expression Omnibus was used for validation. Results A total of 151 differentially expressed protein-coding genes related to survival prognosis were screened out.The machine learning model established with C3orf80,UGP2,and SPC25 demonstrated the best performance. Conclusions Three protein-coding genes (UGP2,C3orf80,and SPC25) were screened out to identify breast invasive carcinoma.This study provides a new direction for the treatment and diagnosis of breast invasive carcinoma.

Lactylation prediction models based on protein sequence and structural feature fusion.

Lysine lactylation (Kla) is a newly discovered posttranslational modification that is involved in important life activities, such as glycolysis-related cell function, macrophage polarization and nervous system regulation, and has received widespread attention due to the Warburg effect in tumor cells. In this work, we first design a natural language processing method to automatically extract the 3D structural features of Kla sites, avoiding potential biases caused by manually designed structural features. Then, we establish two Kla prediction frameworks, Attention-based feature fusion Kla model (ABFF-Kla) and EBFF-Kla, to integrate the sequence features and the structure features based on the attention layer and embedding layer, respectively. The results indicate that ABFF-Kla and Embedding-based feature fusion Kla model (EBFF-Kla), which fuse features from protein sequences and spatial structures, have better predictive performance than that of models that use only sequence features. Our work provides an approach for the automatic extraction of protein structural features, as well as a flexible framework for Kla prediction. The source code and the training data of the ABFF-Kla and the EBFF-Kla are publicly deposited at: https://github.com/ispotato/Lactylation_model.

Global profiling of protein lactylation in Caenorhabditis elegans.

Lactylation, as a novel posttranslational modification, is essential for studying the functions and regulation of proteins in physiological and pathological processes, as well as for gaining in-depth knowledge on the occurrence and development of many diseases, including tumors. However, few studies have examined the protein lactylation of one whole organism. Thus, we studied the lactylation of global proteins in Caenorhabditis elegans to obtain an in vivo lactylome. Using an MS-based platform, we identified 1836 Class I (localization probabilities > 0.75) lactylated sites in 487 proteins. Bioinformatics analysis showed that lactylated proteins were mainly located in the cytoplasm and involved in the tricarboxylic acid cycle (TCA cycle) and other metabolic pathways. Then, we evaluated the conservation of lactylation in different organisms. In total, 41 C. elegans proteins were lactylated and homologous to lactylated proteins in humans and rats. Moreover, lactylation on H4K80 was conserved in three species. An additional 238 lactylated proteins were identified in C. elegans for the first time. This study establishes the first lactylome database in C. elegans and provides a basis for studying the role of lactylation.

Global profiling of lysine lactylation in human lungs.

Lactate is closely related to various cellular processes, such as angiogenesis, responses to hypoxia, and macrophage polarization, while regulating natural immune signaling pathways and promoting neurogenesis and cognitive function. Lysine lactylation (Kla) is a novel posttranslational modification, the examination of which may lead to new understanding of the nonmetabolic functions of lactate and the various physiological and pathological processes in which lactate is involved, such as infection, tumorigenesis and tumor development. Using liquid chromatography-tandem mass spectrometry (LC-MS/MS), researchers have identified lactylation in human gastric cancer cells and some other species, but no research on lactylation in human lungs has been reported. In this study, we performed global profiling of lactylation in human lungs under normal physiological conditions, and 724 Kla sites in 451 proteins were identified. After comparing the identified proteins with those reported in human lactylation datasets, 141 proteins that undergo lactylation were identified for the first time in this study. Our work expands the database on human lactylation and helps advance the study on lactylation function and regulation under physiological and pathological conditions.

Using ATCLSTM-Kcr to predict and generate the human lysine crotonylation database.

Lysine crotonylation (Kcr) is an evolutionarily conserved protein post-translational modifications, which plays an important role in cellular physiology and pathology, such as chromatin remodeling, gene transcription regulation, telomere maintenance, inflammation, and cancer. Tandem mass spectrometry (LC-MS/MS) has been used to identify the global Kcr profiling of human, at the same time, many computing methods have been developed to predict Kcr sites without high experiment cost. Deep learning network solves the problem of manual feature design and selection in traditional machine learning (NLP), especially the algorithms in natural language processing which treated peptides as sentences, thus can extract more in-depth information and obtain higher accuracy. In this work, we establish a Kcr prediction model named ATCLSTM-Kcr which use self-attention mechanism combined with NLP method to highlight the important features and further capture the internal correlation of the features, to realize the feature enhancement and noise reduction modules of the model. Independent tests have proved that ATCLSTM-Kcr has better accuracy and robustness than similar prediction tools. Then, we design pipeline to generate MS-based benchmark dataset to avoid the false negatives caused by MS-detectability and improve the sensitivity of Kcr prediction. Finally, we develop a Human Lysine Crotonylation Database (HLCD) which using ATCLSTM-Kcr and the two representative deep learning models to score all lysine sites of human proteome, and annotate all Kcr sites identified by MS of current published literatures. HLCD provides an integrated platform for human Kcr sites prediction and screening through multiple prediction scores and conditions, and can be accessed on the website:www.urimarker.com/HLCD/. SIGNIFICANCE: Lysine crotonylation (Kcr) plays an important role in cellular physiology and pathology, such as chromatin remodeling, gene transcription regulation and cancer. To better elucidate the molecular mechanisms of crotonylation and reduce the high experimental cost, we establish a deep learning Kcr prediction model and solve the problem of false negatives caused by the detectability of mass spectrometry (MS). Finally, we develop a Human Lysine Crotonylation Database to score all lysine sites of human proteome, and annotate all Kcr sites identified by MS of current published literatures. Our work provides a convenient platform for human Kcr sites prediction and screening through multiple prediction scores and conditions.

[Role of Eukaryotic Translation Elongation Factor 1 Family Members in the Tumorigenesis and Progression of Lung Adenocarcinoma].

Objective To investigate the role and mechanism of eukaryotic translation elongation factor 1(EEF1) family members (EEF1D,EEF1A1,and EEF1A2) in lung adenocarcinoma (LUAD) based on public databases.Methods We examined EEF1 member expression levels in human LUAD samples via The Cancer Genome Atlas in the UCSC Xena browser and the Clinical Proteomic Tumor Analysis Consortium.We analyzed the mRNA and protein levels of EEF1D,EEF1A1,and EEF1A2 and their correlations with pathological variables via the Mann-Whitney U test.The Kaplan-Meier curves were established to assess the prognostic values of EEF1D,EEF1A1,and EEF1A2.The single-sample gene set enrichment analysis algorithm was employed to explore the relationship between the expression levels of EEF1 members and tumor immune cell infiltration.Spearman and Pearson correlation analyses were performed to examine the relationship between the expression levels of EEF1 members and those of the genes in the phosphatidylinositol 3-kinase/protein kinase B signaling pathway.The immunohistochemical assay was employed to determine the expression levels of EEF1D,EEF1A1,and EEF1A2 in the LUAD tissue (n=75) and paracancer tissue (n=75) samples.Results The mRNA and protein levels of EEF1D,EEF1A1,and EEF1A2 showed significant differences between tumor and paracancer tissues (all P<0.001).The patients with high protein levels of EEF1A1 showed bad prognosis in terms of overall survival (P=0.039),and those with high protein levels of EEF1A2 showed good prognosis in terms of overall survival (P=0.012).The influence of the mRNA level of EEF1D on prognosis was associated with pathological characteristics.The expression levels of EEF1 members were significantly associated with the infiltration of various immune cells and the expression of key molecules in the phosphatidylinositol 3-kinase/protein kinase B signaling pathway.Conclusion EEF1D,EEF1A1,and EEF1A2 are associated with the progression of LUAD,serving as the candidate prognostic markers for LUAD.