Integration of machine learning for developing a prognostic signature related to programmed cell death in colorectal cancer

BACKGROUND

Colorectal cancer (CRC) presents a significant global health burden, characterized by a heterogeneous molecular landscape and various genetic and epigenetic alterations. Programmed cell death (PCD) plays a critical role in CRC, offering potential targets for therapy by regulating cell elimination processes that can suppress tumor growth or trigger cancer cell resistance. Understanding the complex interplay between PCD mechanisms and CRC pathogenesis is crucial. This study aims to construct a PCD-related prognostic signature in CRC using machine learning integration, enhancing the precision of CRC prognosis prediction.

METHOD

We retrieved expression data and clinical information from the Cancer Genome Atlas and Gene Expression Omnibus (GEO) datasets. Fifteen forms of PCD were identified, and corresponding gene sets were compiled. Machine learning algorithms, including Lasso, Ridge, Enet, StepCox, survivalSVM, CoxBoost, SuperPC, plsRcox, random survival forest (RSF), and gradient boosting machine, were integrated for model construction. The models were validated using six GEO datasets, and the programmed cell death score (PCDS) was established. Further, the model's effectiveness was compared with 109 transcriptome-based CRC prognostic models.

RESULT

Our integrated model successfully identified differentially expressed PCD-related genes and stratified CRC samples into four subtypes with distinct prognostic implications. The optimal combination of machine learning models, RSF + Ridge, showed superior performance compared with traditional methods. The PCDS effectively stratified patients into high-risk and low-risk groups, with significant survival differences. Further analysis revealed the prognostic relevance of immune cell types and pathways associated with CRC subtypes. The model also identified hub genes and drug sensitivities relevant to CRC prognosis.

CONCLUSION

The current study highlights the potential of integrating machine learning models to enhance the prediction of CRC prognosis. The developed prognostic signature, which is related to PCD, holds promise for personalized and effective therapeutic interventions in CRC.

A novel cuproptosis-related prognostic 2-lncRNAs signature in breast cancer

Background: Cuproptosis, a newly defined regulated form of cell death, is mediated by the accumulation of copper ions in cells and related to protein lipoacylation. Seven genes have been reported as key genes of cuproptosis phenotype. Cuproptosis may be developed by subsequent research as a target to treat cancer, such as breast cancer. Long-noncoding RNA (lncRNA) has been proved to play a vital role in regulating the biological process of breast cancer. However, the role of lncRNAs in cuproptosis is poorly studied. Methods: Based on TCGA (The Cancer Genome Atlas) database and integrated several R packages, we screened out 153 cuproptosis-related lncRNAs and constructed a novel cuproptosis-related prognostic 2-lncRNAs signature (BCCuS) in breast cancer and then verified. By using pRRophetic package and machine learning, 72 anticancer drugs, significantly related to the model, were screened out. qPCR was used to detect the differentially expression of two model lncRNAs and seven cuproptosis genes between 10 pairs of breast cancer tissue samples and adjacent samples. Results: We constructed a novel cuproptosis-related prognostic 2-lncRNAs (USP2-AS1, NIFK-AS1) signature (BCCuS) in breast cancer. Univariate COX analysis (p < .001) and multivariate COX analysis (p < .001) validated that BCCuS was an independent prognostic factor for breast cancer. Overall survival Kaplan Meier-plotter, ROC curve and Risk Plot validated the prognostic value of BCCuS both in test set and verification set. Nomogram and C-index proved that BCCuS has strong correlation with clinical decision-making. BCCuS still maintain inspection efficiency when patients were splitting into Stage I-II (p = .024) and Stage III-IV (p = .003) breast cancer. BCCuS-high group and BCCuS-low group showed significant differences in gene mutation frequency, immune function, TIDE (tumor immune dysfunction and exclusion) score and other phenotypes. TMB (tumor mutation burden)-high along with BCCuS-high group had the lowest Survival probability (p = .005). 36 anticancer drugs whose sensitivity (IC50) was significantly related to the model were screened out using pRRophetic package. qPCR results showed that two model lncRNAs (USP2-AS1, NIFK-AS1) and three Cuproptosis genes (FDX1, PDHA1, DLAT) expressed differently between 10 pairs of breast cancer tissue samples and adjacent samples. Conclusion: The current study reveals that cuproptosis-related prognostic 2-lncRNAs signature (BCCuS) may be useful in predicting the prognosis, biological characteristics, and appropriate treatment of breast cancer patients.

[Application value of combined deep inhalation and breath-hold technique guided by four-dimensional CT in preoperative localization of solitary pulmonary nodules]

Objective: To explore the clinical value of four dimensional computed tomography (4 D CT) guided combined with deep inhalation and breath hold (DIBH) technique in the preoperative localization of solitary pulmonary nodules. Methods: The data of a total of 106 patients with solitary pulmonary nodules from March 2018 to May 2021 in the Ningbo First Hospital were collected retrospectively. Among them, there were 26 males and 80 females aged from 21 to 83 (47.4±14.2) years. According to different localization methods, 53 cases were divided into the control group, as the pulmonary nodules were located by CT guided injection of indocyanine green under calm breathing and 53 cases were divided into in the experimental group, as those patients were treated with indocyanine green injection under the guidance of 4 D CT combined with DIBH technology to locate pulmonary nodules. The three-dimensional distance deviation between pulmonary nodules and indocyanine green injection points was compared between the two groups to obtain the accuracy of pulmonary nodule localization. The preoperative positioning time of the two groups was compared by timing. Results: Among the 106 patients, there were 46 pure ground glass nodules, 32 sub solid nodules and 28 solid nodules, all of which were successfully localized before operation, with a success rate of 100%. The size of pulmonary nodules in the control group was (9.1±2.3) mm and the three-dimensional deviation[M(Q₁, Q₃)]between indocyanine green injection site and pulmonary nodules was X axis [7.0 (3.7, 12.6)] mm, Y axis [6.6 (2.9, 11.2)] mm, Z axis [3.0 (2.0, 6.0)]mm, respectively, and the preoperative positioning time was (11.4±3.8) min. The size of pulmonary nodules in the experimental group was (8.9±2.1) mm, and the deviations in 3 D direction were X axis [4.8 (3.0, 7.9)]mm, Y axis [3.8 (1.3, 7.5)]mm, Z axis [4.0 (2.0, 6.0)] mm, respectively. The preoperative positioning time was (9.3±3.0) min. There were statistically significant differences in preoperative positioning time and deviation of X and Y axis between the experimental group and the control group (P<0.05), but no statistically significant differences was found in deviation of Z axis (P>0.05). Conclusion: 4 D CT guided DIBH technology could improve the accuracy of preoperative localization of pulmonary nodules and save operation time, which is worthy of popularization.

RBMS2 Chemosensitizes Breast Cancer Cells to Doxorubicin by Regulating BMF Expression

Chemoresistance is closely related to the therapeutic effect and prognosis in breast cancer patients. Increasing evidences demonstrated that RNA binding proteins (RBPs) have notable roles in regulating cancer cell proliferation, metastasis and chemotherapeutic sensitivity. RNA binding motif single stranded interacting protein 2 (RBMS2), an RBP, has been considered to be a tumor suppressor in several cancers. However, its role of doxorubicin sensitivity in breast cancer patients has not yet been fully revealed. Here, we performed doxorubicin cytotoxicity assay, flow cytometry and mouse xenograft model to examine the influence of RBMS2 on doxorubicin sensitization in vitro and in vivo. RIP assay and dual-luciferase reporter assay were performed to explore the relationship between RBMS2 and BMF. Our data demonstrated that upregulation of RBMS2 in breast cancer cells could enhance sensitivity to doxorubicin and promote apoptosis in the presence of doxorubicin, while inhibition of RBMS2 showed an opposite trend. Moreover, this chemosensitizing effect of RBMS2 could be reversed by the inhibition of Bcl-2 modifying factor (BMF). RBMS2 positively regulated BMF expression and increased BMF-induced expression of (cleaved) caspase 3, (cleaved) caspase 9 and poly (ADP-Ribose) polymerase (PARP). These results uncovered a novel mechanism for RBMS2 in the sensibilization of doxorubicin, suggesting that RBMS2 may act as a potential therapeutic target for drug-resistant breast cancer.

Process of aerenchyma formation and reactive oxygen species induced by waterlogging in wheat seminal roots

The development and regulation of aerenchyma in waterlogged conditions were studied in the seminal roots of wheat. Evans blue staining and the first cell death position indicated that the cortical cell death began at the root mid-cortex cells in flooding conditions. Continuous waterlogging treatment caused the spread of cell death from the mid-cortex to the neighboring cells and well-developed aerenchyma was formed after 72 h. Meanwhile, the formation of radial oxygen loss barrier was observed in the exodermis owing to the induction of Casparian bands and lignin deposition. Analysis of aerenchyma along the wheat root revealed that aerenchyma formed at 10 mm from the root tip, significantly increased toward the center of the roots, and decreased toward the basal region of the root. In situ detection of radial oxygen species (ROS) showed that ROS accumulation started in the mid-cortex cells, where cell death began indicating that cell death was probably accompanied by ROS production. Further waterlogging treatments resulted in the accumulation of ROS in the cortical cells, which were the zone for aerenchyma development. Accumulation and distribution of H₂O₂ at the subcellular level were revealed by ultracytochemical localization, which further verified the involvement of ROS in the cortical cell death process (i.e., aerenchyma formation). Furthermore, gene expression analysis indicated that ROS production might be the result of up-regulation of genes encoding for ROS-producing enzymes and the down-regulation of genes encoding for ROS-detoxifying enzymes. These results suggest that aerenchyma development in wheat roots starts in the mid-cortex cells and its formation is regulated by ROS.

Cell wall degradation and the dynamic changes of Ca2+ and related enzymes in the developing aerenchyma of wheat (Triticum aestivum L.) under waterlogging

This research was aimed to study the cell wall degradation and the dynamic changes of Ca2+ and related enzymes in developing aerenchyma of wheat root under waterlogging. An examination of morphological development by light and electron microscope revealed that the structure of cell wall in middle cortical cells remained intact after 12 h of waterlogging and turned thinner after waterlogging for 24 h. At 48 h, the aerenchyma has been formed. The cellulase activity gradually increased in middle cortical cells within 24 h of waterlogging, and decreased with the formation of aerenchyma. Fluorescence detection and subcellular localization of Ca2+ showed the dynamic changing of Ca2+ at the cellular and subcellular levels during the development of aerenchyma. The activity of Ca2+-ATPase enhanced markedly in intercellular space, plasma membrane and tonoplast of some middle cortical cells after 8 h of waterlogging and remained high after 24 h, but it decreased after 48 h of waterlogging. All these suggests that cellulase, Ca2+ and Ca2+-ATPase show a dynamic distribution during the aerenchyma development which associated with the cell wall degradation of middle cortical cells. Moreover, there is a feedback regulation between Ca2+ and Ca2+-ATPase.

[Ventricular arrhythmias in patients with cor pulmonale]

This paper reported the results of 24 hours continuous ECG, day time and overnight arterial blood gas/pH and serum electrolytes in 20 patients with cor pulmonale, in order to investigate the rule of changes of ventricular arrhythmias (VA). The results were as follows: (1) Incidence of VA in 24 hours Holter Monitoring was 100%. (2) Frequent VPBs were increased significantly in period of acute attack (40%, P less than 0.05). (3) Non-sustained VT in the nocturnal (40%) was more than the day time in period of attack (P less than 0.05). (4) Relationship between VA and variables in period of attack: PaCO2 (r = 0.90) greater than arterial blood pH (r = -0.87) greater than PaO2 (r = -0.63); in relieved period VA were only related to serum potassium (r = -0.86). The authors speculated the severity of VA in period of attack was related with worse of arterial blood gas/pH, cardiac dysfunction and compensated enhance of sympathetic activity. It seems that the view accorded with theory of Yin-Yang in TCM.

Partial amino acid sequence of apolipoprotein(a) shows that it is homologous to plasminogen

Apolipoprotein(a) [apo(a)] is a glycoprotein with Mr approximately equal to 280,000 that is disulfide linked to apolipoprotein B in lipoprotein(a) particles. Elevated plasma levels of lipoprotein(a) are correlated with atherosclerosis. Partial amino acid sequence of apo(a) shows that it has striking homology to plasminogen. Plasminogen is a plasma serine protease zymogen that consists of five homologous and tandemly repeated domains called kringles and a trypsin-like protease domain. The amino-terminal sequence obtained for apo(a) is homologous to the beginning of kringle 4 but not the amino terminus of plasminogen. Apo(a) was subjected to limited proteolysis by trypsin or V8 protease, and fragments generated were isolated and sequenced. Sequences obtained from several of these fragments are highly (77-100%) homologous to plasminogen residues 391-421, which reside within kringle 4. Analysis of these internal apo(a) sequences revealed that apo(a) may contain at least two kringle 4-like domains. A sequence obtained from another tryptic fragment also shows homology to the end of kringle 4 and the beginning of kringle 5. Sequence data obtained from two tryptic fragments show homology with the protease domain of plasminogen. One of these sequences is homologous to the sequences surrounding the activation site of plasminogen. Plasminogen is activated by the cleavage of a specific arginine residue by urokinase and tissue plasminogen activator; however, the corresponding site in apo(a) is a serine that would not be cleaved by tissue plasminogen activator or urokinase. Using a plasmin-specific assay, no proteolytic activity could be demonstrated for lipoprotein(a) particles. These results suggest that apo(a) contains kringle-like domains and an inactive protease domain.