Lung cancer postoperative complication prediction (PCP) is significant for decreasing the perioperative mortality rate after lung cancer surgery. In this paper we concentrate on two PCP tasks: (1) the binary classification for predicting whether a patient will have postoperative complications; and (2) the three-class multi-label classification for predicting which postoperative complication a patient will experience. Furthermore, an important clinical requirement of PCP is the extraction of crucial variables from electronic medical records. We propose a novel multi-layer perceptron (MLP) model called medical MLP (MediMLP) together with the gradient-weighted class activation mapping (Grad-CAM) algorithm for lung cancer PCP. The proposed MediMLP, which involves one locally connected layer and fully connected layers with a shortcut connection, simultaneously extracts crucial variables and performs PCP tasks. The experimental results indicated that MediMLP outperformed normal MLP on two PCP tasks and had comparable performance with existing feature selection methods. Using MediMLP and further experimental analysis, we found that the variable of "time of indwelling drainage tube" was very relevant to lung cancer postoperative complications.

Background BSG (CD147) is a member of the immunoglobulin superfamily that shows roles for potential prognostics and therapeutics for metastatic cancers and SARS-CoV-2 invasion for COVID-19. The susceptibility of malignant cancers to SARS-CoV-2 as well as the correlations between disease outcome and BSG expression in tumor tissues have not been studied in depth. Methods In this study, we explored the BSG expression profile, survival correlation, DNA methylation, mutation, diagnostics, prognostics, and tumor-infiltrating lymphocytes (TILs) from different types of cancer tissues with corresponding healthy tissues. In vitro studies for cordycepin (CD), N6-(2-hydroxyethyl) adenosine (HEA), N6, N6-dimethyladenosine (m⁶₂A) and 5′-uridylic acid (UMP) on BSG expression were also conducted. Results We revealed that BSG is conserved among different species, and significantly upregulated in seven tumor types, including ACC, ESCA, KICH, LIHC, PAAD, SKCM and THYM, compared with matched normal tissues, highlighting the susceptibility of these cancer patients to SARS-CoV-2 invasion, COVID-19 severity and progression of malignant cancers. High expression in BSG was significantly correlated with a short OS in LGG, LIHC and OV patients, but a long OS in KIRP patients. Methylation statuses in the BSG promoter were significantly higher in BRCA, HNSC, KIRC, KIRP, LUSC, PAAD, and PRAD tumor tissues, but lower in READ. Four CpGs in the BSG genome were identified as potential DNA methylation biomarkers which could be used to predict malignant cancers from normal individuals. Furthermore, a total of 65 mutation types were found, in which SARC showed the highest mutation frequency (7.84%) and THYM the lowest (0.2%). Surprisingly, both for disease-free and progression-free survival in pan-cancers were significantly reduced after BSG mutations. Additionally, a correlation between BSG expression and immune lymphocytes of CD56bright natural killer cell, CD56dim natural killer cell and monocytes, MHC molecules of HLA-A, HLA-B, HLA-C and TAPBP, immunoinhibitor of PVR, PVRL2, and immunostimulators of TNFRSF14, TNFRSF18, TNFRSF25, and TNFSF9, was revealed in most cancer types. Moreover, BSG expression was downregulated by CD, HEA, m⁶₂A or UMP in cancer cell lines, suggesting therapeutic potentials for interfering entry of SARS-CoV-2. Conclusions Altogether, our study highlights the values of targeting BSG for diagnostic, prognostic and therapeutic strategies to fight malignant cancers and COVID-19. Small molecules CD, HEA, m⁶₂A and UMP imply therapeutic potentials in interfering with entry of SARS-CoV-2 and progression of malignant cancers.