Xi Zhang

Neoantigens have emerged as promising targets for personalized cancer immunotherapy. However, accurate identification of immunogenic neoantigens remains a challenge due to limitations in existing predictive models. Here, we present CNNeo, a novel deep learning-based neoantigen prediction model, and CNNeoPP, an integrated computational pipeline for neoantigen discovery. CNNeo employs large language model-derived sequence representations and multi-modal feature integration, demonstrating superior predictive performance compared to existing tools. CNNeoPP was rigorously validated using independent datasets, including the TESLA dataset, and experimental validation via ELISpot T-cell assays. Additionally, we conducted a proof-of-concept study utilizing plasma cell-free DNA to explore the feasibility of non-invasive neoantigen prediction. We found that increased sequencing depth enhances neoantigen detectability, further amplified by the prioritization strategy of CNNeoPP. CNNeoDB, a publicly accessible database was developed compiling neoantigen data from multiple sources. This study establishes robust tools for neoantigen prediction, with implications for optimizing cancer immunotherapy and liquid biopsy-based tumor monitoring. CNNeoPP is available at https://github.com/AaronChen007/neoantigen.

Neoantigens have emerged as promising targets for personalized cancer immunotherapy. However, accurate identification of immunogenic neoantigens remains a challenge due to limitations in existing predictive models. Here, we present CNNeo, a novel deep learning-based neoantigen prediction model, and CNNeoPP, an integrated computational pipeline for neoantigen discovery. CNNeo employs natural language processing-based sequence encoding and multi-modal feature integration, demonstrating superior predictive performance compared to existing tools. CNNeoPP was rigorously validated using independent datasets, including the TESLA dataset, and experimental validation via ELISpot T-cell assays. Additionally, we conducted a proof-of-concept study utilizing plasma cell-free DNA to explore the feasibility of non-invasive neoantigen prediction. We found that increased sequencing depth enhances neoantigen detectability, further amplified by by the prioritization strategy of CNNeoPP. CNNeoDB, a publicly accessible database was developed compiling neoantigen data from multiple sources. This study establishes robust tools for neoantigen prediction, with implications for optimizing cancer immunotherapy and liquid biopsy-based tumor monitoring.

Breast cancer (BC) subtypes exhibit distinct epigenetic landscapes, with triple-negative breast cancer (TNBC) lacking effective targeted therapies. This study investigates histone biomarkers and therapeutic vulnerabilities across BC subtypes. The immunohistochemical profiling of >20 histone biomarkers, including histone modifications, modifiers, and oncohistone mutations, was conducted on a discovery cohort and a validation cohort of BC tissues, healthy controls, and cell line models. Transcriptomic and cell growth analyses were conducted to evaluate the effects of the small-molecule G9a inhibitor in diverse BC models. Key histone biomarkers, including H3K9me2, H3K36me2, and H3K79me, were differentially expressed across BC subtypes. H3K9me2 emerged as an independent predictor for distinguishing TNBC from other less-aggressive BC subtypes, with elevated expression correlating with higher tumor grade and stage. G9a inhibition impaired cell proliferation and modulated epithelial–mesenchymal transition pathways, with the strongest impact in basal-like TNBC. The disruption of the oncogene and tumor suppressor regulation (e.g., TP53, SATB1) was observed in TNBC. This study highlights G9a’s context-dependent roles in BC, supporting its potential as a therapeutic target. The findings provide a foundation for subtype-specific epigenetic therapies to improve outcomes in aggressive BC subtypes.

Background: Cerebrospinal fluid (CSF) culture is the diagnostic gold standard for neuroinfectious diseases such as bacterial meningitis, but its sensitivity is limited and results are often delayed. Natural language processing (NLP) offers a powerful approach to extract meaningful clinical signals from unstructured data such as chief complaints and ICD notes. This study applies machine learning, including BioBERT-enhanced NLP models (not traditional TF-IDF approaches), to support early diagnosis and outcome prediction in ICU patients. Methods: Training and validation datasets were derived from MIMIC-IV (internal) and MIMIC-III/eICU (external) databases. Fully connected neural network (FCNN) and other machine learning models were trained to predict CSF culture results using structured lab features. Labels were refined using clinical criteria to reduce false negatives. For ICU survival prediction, three multimodal deep learning architectures (mCNN, mFCNN, and mLSTM) were developed using two ICU survival cohorts with different inclusion criteria. The Strict ICU Survival Cohort included CSF culture results as an input feature, while the Lenient ICU Survival Cohort excluded this requirement, allowing for a broader patient population. In both cohorts, models integrated structured variables with unstructured text encoded by BioBERT, a deep contextual language model, rather than simpler methods like TF-IDF, effectively capturing clinical meaning from free-text ICD entries and chief complaints. Results: For CSF culture prediction (training n = 9261), the FCNN model achieved the highest performance (AUROC = 0.853) in independent validation. For ICU survival prediction in the Strict ICU Survival Cohort (training n = 5,795), the mCNN model achieved an AUROC of 0.889 in external validation. In the expanded Lenient ICU Survival Cohort (n = 58,615), the same model achieved an AUROC of 0.974 and an AUPRC of 0.868 during external validation. During model training and development, the predictive performance declined when text features were excluded (AUROC from 0983 to 0.946) or when ICD entries were converted from free-text (BioBERT-encoded) to coded format (AUROC to 0.947). Conclusions: Multimodal machine learning models, enhanced by advanced NLP through BioBERT embeddings of clinical free text, effectively predicted CSF culture results and ICU survival outcomes.

Background: Breast cancer (BC) subtypes exhibit distinct epigenetic landscapes, with triple-negative breast cancer (TNBC) lacking effective targeted therapies. This study investigates histone biomarkers and therapeutic vulnerabilities across BC subtypes. Methods: Immunohistochemical profiling of >20 histone biomarkers, including histone modifications, modifiers and oncohistone mutations was conducted on a discovery cohort and a validation cohort of BC tissues, healthy controls and cell line models. Transcriptomic and cell growth analyses were conducted to evaluate the effects of the small-molecule G9a inhibitor in diverse BC models. Results: Key histone biomarkers, including H3K9me2, H3K36me2, and H3K79me, were differentially expressed across BC subtypes. H3K9me2 emerged as an independent predictor for distinguishing TNBC from other less aggressive BC subtypes, with elevated expression correlating with higher tumor grade and stage. G9a inhibition impaired cell proliferation and modulated epithelial-mesenchymal transition pathways, with the strongest impact in basal-like TNBC. Disruption of oncogene and tumor suppressor regulation (e.g., TP53, SATB1) was observed in TNBC. Conclusion: This study highlights G9a’s context-dependent roles in BC, supporting its potential as a therapeutic target. Findings provide a foundation for subtype-specific epigenetic therapies to improve outcomes in aggressive BC subtypes. Clinical Perspective: This study was undertaken to explore the epigenetic landscape of breast cancer subtypes, focusing on histone modifications and their therapeutic potential, particularly through targeting G9a, a histone methyltransferase. The study identified subtype-specific histone biomarkers, with H3K9me2 emerging as a key marker distinguishing TNBC from other less aggressive subtypes. G9a inhibition demonstrated robust anti-cancer effects, including cell proliferation impairment and disruption of oncogenic pathways, particularly in TNBC models. These findings provide compelling evidence for the development of subtype-specific epigenetic therapies targeting G9a and similar regulators, highlighting their potential to reshape the treatment landscape for TNBC and improve patient outcomes.

Background: Despite the improved survival observed in PD-1/PD-L1 blockade therapy, a substantial proportion of cancer patients, including those with non-small cell lung cancer (NSCLC), still lack a response. Methods: Transcriptomic profiling was conducted on a discovery cohort comprising 100 whole blood samples, as collected multiple times from 48 healthy controls (including 43 published data) and 31 NSCLC patients that under treatment with a combination of anti-PD-1 Tislelizumab and chemotherapy. Differentially expressed genes (DEGs), simulated immune cell subsets, and germline DNA mutational markers were identified from patients achieved a pathological complete response during the early treatment cycles. The predictive values of mutational markers were further validated in an independent immunotherapy cohort of 1661 subjects, and then confirmed in genetically matched lung cancer cell lines by a co-culturing model. Results: The gene expression of hundreds of DEGs (FDR p < 0.05, fold change < -2 or > 2) distinguished responders from healthy controls, indicating the potential to stratify patients utilizing early on-treatment features from blood. PD-1-mediated cell abundance changes in memory CD4 + and regulatory T cell subset were more significant or exclusively observed in responders. A panel of top-ranked genetic alterations showed significant associations with improved survival (p < 0.05) and heightened responsiveness to anti-PD-1 treatment in patient cohort and co-cultured cell lines. Conclusion: This study discovered and validated peripheral blood-based biomarkers with evident predictive efficacy for early therapy response and patient stratification before treatment for neoadjuvant PD-1 blockade in NSCLC patients.

The field of cancer neoantigen investigation has developed swiftly in the past decade. Predicting novel and true neoantigens derived from large multi-omics data became difficult but critical challenges. The rise of Artificial Intelligence (AI) or Machine Learning (ML) in biomedicine application has brought benefits to strengthen the current computational pipeline for neoantigen prediction. ML algorithms offer powerful tools to recognize the multidimensional nature of the omics data and therefore extract the key neoantigen features enabling a successful discovery of new neoantigens. The present review aims to outline the significant technology progress of machine learning approaches, especially the newly deep learning tools and pipelines, that were recently applied in neoantigen prediction. In this review article, we summarize the current state-of-the-art tools developed to predict neoantigens. The standard workflow includes calling genetic variants in paired tumor and blood samples, and rating the binding affinity between mutated peptide, MHC (I and II) and T cell receptor (TCR), followed by characterizing the immunogenicity of tumor epitopes. More specifically, we highlight the outstanding feature extraction tools and multi-layer neural network architectures in typical ML models. It is noted that more integrated neoantigen-predicting pipelines are constructed with hybrid or combined ML algorithms instead of conventional machine learning models. In addition, the trends and challenges in further optimizing and integrating the existing pipelines are discussed.

Cystic fibrosis (CF) is an inherited disorder caused by biallelic mutations of the CF transmembrane conductance regulator ( CFTR ) gene. Converging evidence suggests that CF carriers with only 1 defective CFTR copy are at increased risk for CF-related conditions and pulmonary infections, but the molecular mechanisms underpinning this effect remain unknown. We performed transcriptomic profiling of peripheral blood mononuclear cells (PBMCs) of CF child-parent trios (proband, father, and mother) and healthy control (HC) PBMCs or THP-1 cells incubated with the plasma of these participants. Transcriptomic analyses revealed suppression of cytokine-enriched immune-related genes ( IL-1 β , CXCL8 , CREM ), implicating lipopolysaccharide tolerance in innate immune cells (monocytes) of CF probands and their parents. These data suggest that a homozygous as well as a heterozygous CFTR mutation can modulate the immune/inflammatory system. This conclusion is further supported by the finding of lower numbers of circulating monocytes in CF probands and their parents, compared with HCs, and the abundance of mononuclear phagocyte subsets, which correlated with Pseudomonas aeruginosa infection, lung disease severity, and CF progression in the probands. This study provides insight into demonstrated CFTR -related innate immune dysfunction in individuals with CF and carriers of a CFTR mutation that may serve as a target for personalized therapy.

A massive number of transcriptomic profiles of blood samples from COVID-19 patients has been produced since pandemic COVID-19 begins, however, these big data from primary studies have not been well integrated by machine learning approaches. Taking advantage of modern machine learning arthrograms, we integrated and collected single cell RNA-seq (scRNA-seq) data from three independent studies, identified genes potentially available for interpretation of severity, and developed a high-performance deep learning-based deconvolution model AImmune that can predict the proportion of seven different immune cells from the bulk RNA-seq results of human peripheral mononuclear cells. This novel approach which can be used for clinical blood testing of COVID-19 on the ground that previous research shows that mRNA alternations in blood-derived PBMCs may serve as a severity indicator. Assessed on real-world data sets, the AImmune model outperformed the most recognized immune profiling model CIBERSORTx. The presented study showed the results obtained by the true scRNA-seq route can be consistently reproduced through the new approach AImmune, indicating a potential replacing the costly scRNA-seq technique for the analysis of circulating blood cells for both clinical and research purposes.