Predicting cell types with supervised contrastive learning on cells and their types.

Single-cell RNA-sequencing (scRNA-seq) is a powerful technique that provides high-resolution expression profiling of individual cells. It significantly advances our understanding of cellular diversity and function. Despite its potential, the analysis of scRNA-seq data poses considerable challenges related to multicollinearity, data imbalance, and batch effect. One of the pivotal tasks in single-cell data analysis is cell type annotation, which classifies cells into discrete types based on their gene expression profiles. In this work, we propose a novel modeling formalism for cell type annotation with a supervised contrastive learning method, named SCLSC (Supervised Contrastive Learning for Single Cell). Different from the previous usage of contrastive learning in single cell data analysis, we employed the contrastive learning for instance-type pairs instead of instance-instance pairs. More specifically, in the cell type annotation task, the contrastive learning is applied to learn cell and cell type representation that render cells of the same type to be clustered in the new embedding space. Through this approach, the knowledge derived from annotated cells is transferred to the feature representation for scRNA-seq data. The whole training process becomes more efficient when conducting contrastive learning for cell and their types. Our experiment results demonstrate that the proposed SCLSC method consistently achieves superior accuracy in predicting cell types compared to five state-of-the-art methods. SCLSC also performs well in identifying cell types in different batch groups. The simplicity of our method allows for scalability, making it suitable for analyzing datasets with a large number of cells. In a real-world application of SCLSC to monitor the dynamics of immune cell subpopulations over time, SCLSC demonstrates a capability to discriminate cell subtypes of CD19+ B cells that were not present in the training dataset.

The transcriptome signature analysis of the epithelial-mesenchymal transition and immune cell infiltration in colon adenocarcinoma.

The epithelial-mesenchymal transition (EMT) process is tightly connected to tumors' immune microenvironment. In colon adenocarcinoma (COAD), both the EMT and immune cell infiltration contribute to tumor progression; however, several questions regarding the mechanisms governing the interaction between EMT and the immune response remain unanswered. Our study aims to investigate the cross-talk between these two processes in cases of COAD and identify the key regulators involved. We utilized the EMT and immune signatures of samples from the COAD-TCGA database to identify three subtypes of COAD: high mesenchymal, medium mesenchymal, and low mesenchymal. We observed that EMT was associated with increased tumor immune response and infiltration mediated by pro-inflammatory cytokines. However, EMT was also linked to immunosuppressive activity that involved regulatory T cells, dendritic cells, and the upregulated expression of multiple immune checkpoints, such as PD-1, PDL-1, CTLA-4, and others. Finally, we employed the multivariate random forest feature importance method to identify key genes, such as DOK2 and MSRB3, that may play crucial roles in both EMT and the intratumoral immune response.

Identifying Key Regulators of Keratinization in Lung Squamous Cell Cancer Using Integrated TCGA Analysis.

Keratinization is one of lung squamous cell cancer's (LUSC) hallmark histopathology features. Epithelial cells produce keratin to protect their integrity from external harmful substances. In addition to their roles as cell protectors, recent studies have shown that keratins have important roles in regulating either normal cell or tumor cell functions. The objective of this study is to identify the genes and microRNAs (miRNAs) that act as key regulators of the keratinization process in LUSC. To address this goal, we classified LUSC samples from GDC-TCGA databases based on their keratinization molecular signatures. Then, we performed differential analyses of genes, methylation, and miRNA expression between high keratinization and low keratinization samples. By reconstruction and analysis of the differentially expressed genes (DEGs) network, we found that TP63 and SOX2 were the hub genes that were highly connected to other genes and displayed significant correlations with several keratin genes. Methylation analysis showed that the P63, P73, and P53 DNA-binding motif sites were significantly enriched for differentially methylated probes. We identified SNAI2, GRHL3, TP63, ZNF750, and FOXE1 as the top transcription factors associated with these binding sites. Finally, we identified 12 miRNAs that influence the keratinization process by using miRNA-mRNA correlation analysis.

Analyzing integrated network of methylation and gene expression profiles in lung squamous cell carcinoma.

Gene expression, DNA methylation, and their organizational relationships are commonly altered in lung squamous cell carcinoma (LUSC). To elucidate these complex interactions, we reconstructed a differentially expressed gene network and a differentially methylated cytosine (DMC) network by partial information decomposition and an inverse correlation algorithm, respectively. Then, we performed graph union to integrate the networks. Community detection and enrichment analysis of the integrated network revealed close interactions between the cell cycle, keratinization, immune system, and xenobiotic metabolism gene sets in LUSC. DMC analysis showed that hypomethylation targeted the gene sets responsible for cell cycle, keratinization, and NRF2 pathways. On the other hand, hypermethylated genes affected circulatory system development, the immune system, extracellular matrix organization, and cilium organization. By centrality measurement, we identified NCAPG2, PSMG3, and FADD as hub genes that were highly connected to other nodes and might play important roles in LUSC gene dysregulation. We also found that the genes with high betweenness centrality are more likely to affect patients' survival than those with low betweenness centrality. These results showed that the integrated network analysis enabled us to obtain a global view of the interactions and regulations in LUSC.

Integrated analysis of cell shape and movement in moving frame.

The cell's movement and morphological change are two interrelated cellular processes. An integrated analysis is needed to explore the relationship between them. However, it has been challenging to investigate them as a whole. The cell's trajectory can be described by its speed, curvature, and torsion. On the other hand, the three-dimensional (3D) cell shape can be studied by using a shape descriptor such as spherical harmonic (SH) descriptor, which is an extension of a Fourier transform in 3D space. We propose a novel method using parallel-transport (PT) to integrate these shape-movement data by using moving frames as the 3D-shape coordinate system. This moving frame is purely determined by the velocity vector. On this moving frame, the movement change will influence the coordinate system for shape analysis. By analyzing the change of the SH coefficients over time in the moving frame, we can observe the relationship between shape and movement. We illustrate the application of our approach using simulated and real datasets in this paper.

Immuno-PET imaging for non-invasive assessment of cetuximab accumulation in non-small cell lung cancer.

BACKGROUNDS: Overexpression of epidermal growth factor receptor (EGFR) has been established as a valid therapeutic target of non-small cell lung cancer (NSCLC). However, the clinical benefit of cetuximab as an EGFR-targeting drug is still controversial, partially due to the lack of effective means to identify suitable patients. This study aimed to investigate the potential of radiolabeled cetuximab as a non-invasive tool to predict cetuximab accumulation in NSCLC tumor xenografts with varying EGFR expression levels. METHODS: The NSCLC tumors in model mice were subjected to in vivo biodistribution study and positron emission tomography (PET) imaging 48 h after injection of either 111In- or 64Cu-labeled cetuximab. The EGFR expression levels of NSCLC tumors were determined by ex vivo immunoblotting. RESULTS: We found that tumors with high EGFR expression had significantly higher [111In]In-DOTA-cetuximab accumulation than tumors with moderate to low EGFR expression (P < 0.05). Strong correlations were found between [111In]In-DOTA-cetuximab tumor uptake and EGFR expression level (r = 0.893), and between [64Cu]Cu-DOTA-cetuximab tumor uptake with EGFR expression level (r = 0.915). PET imaging with [64Cu]Cu-DOTA-cetuximab allowed clear visualization of tumors. CONCLUSION: Our findings suggest that this immuno-PET imaging can be clinically translated as a tool to predict cetuximab accumulation in NSCLC cancer patients prior to cetuximab therapy.

Applying near-infrared photoimmunotherapy to B-cell lymphoma: comparative evaluation with radioimmunotherapy in tumor xenografts.

OBJECTIVE: Radioimmunotherapy (RIT) has proven effective for patients with relapsed and refractory lymphoma. However, new types of therapy are strongly desired as B-cell lymphoma remains incurable for many patients. Photoimmunotherapy (PIT) is an emerging targeted cancer therapy that uses photosensitizer (IR700)-conjugated monoclonal antibodies (mAbs) to specifically kill cancer cells. To evaluate the usefulness and potential role of PIT for treating B-cell lymphoma in a comparison with RIT, we performed in vivo PIT and RIT studies with an IR700 or 90Y-conjugated anti-CD20 mAb, NuB2. METHODS: IR700 or 90Y were conjugated to NuB2. Since cell aggressiveness greatly affects the therapeutic effect, we selected both an indolent (RPMI 1788) and an aggressive (Ramos) type of B-cell lymphoma cell line. The in vitro therapeutic effect of PIT and the biodistribution profiles of IR700-NuB2 were evaluated. In vivo PIT and RIT studies were performed with 100 or 500 µg of IR700-NuB2 and 150 µCi/20 µg of 90Y-NuB2, respectively, in two types of B-cell lymphoma-bearing mice. RESULTS: The in vitro studies revealed that Ramos was more sensitive than RPMI 1788 to PIT. The therapeutic effect of PIT with 500 µg IR700-NuB2 was superior to any other therapies against aggressive Ramos tumors, whereas RIT showed the highest therapeutic effect in indolent RPMI 1788 tumors. Since the uptake levels and intratumoral distribution of IR700-NuB2 were comparable in both tumors, a possible cause of this difference is the tumor growth rate. The PIT with 500 µg (IR700-NuB2) group showed a significantly greater therapeutic effect than the PIT with 100 µg group due to the higher and more homogeneous tumor distribution of IR700-NuB2. CONCLUSIONS: PIT was effective for both indolent and aggressive B-cell lymphoma, and the higher dose provided a better therapeutic effect. In aggressive tumors, PIT was more effective than RIT. Thus, PIT would be a promising strategy for the locoregional treatment or control of B-cell lymphoma. Since PIT and RIT have distinctive advantages over each other, they could play complementary rather than competitive roles in B-cell lymphoma treatment.