Study Of Keloid Severity And Recurrence Prediction Models Based On Clinical And Bioinformatics Data

Objectives:1.To explore the prediction model and nomogram establishment of recurrence in keloid patients after surgery and radiotherapy.2.To establish,evaluate and compare three recurrence prediction models for keloid patients based on machine learning methods.3.To investigate the differentially expressed genes associated with keloid inflammation at different severity levels and identify potential hub genes.4.To improve the keloid subepidermal vascular network flap(KSVNF)design based on the analysis of recurrence-related factors,and to explore the flap’s best length/width ratio.Methods:1.We collected electronic medical record data on keloid patients at Peking Union Medical College Hospital between 2015 and 2019 and conducted follow-ups to assess recurrence within a 2-year period.After excluding patients with lost follow-up or missing data,we used the data set of chest keloid patients as the training set to develop a logistic regression model and nomogram.These were then tested in a separate data set of non-chest keloid patients.2.We enrolled 301 keloid patients who received surgery and postoperative radiotherapy and divided them into a training set(70%)and a validation set(30%).Three recurrence prediction models were established in the training set:the logistic regression model,the decision tree model,and the random forest model.We then evaluated and compared the performance of these models in the validation set,using metrics such as accuracy,sensitivity,specificity,recall,precision,kappa coefficient,and the area under the ROC curve(AUC).3.We obtained 6 severe keloid samples(defined as KAAS>7)and 6 mild keloid samples(defined as KAAS<4)and conducted high-throughput sequencing using the Oncomine Immune Response Research Assay kit.The resulting sequencing data were analyzed through bioinformatics to identify differentially expressed genes(DEGs)between the two groups and to explore hub genes.To further clarify the DEGs’ main functions and related pathways,we performed gene ontology(GO)annotation and KEGG pathway enrichment analysis on these genes.4.A total of 35 KSVNFs were designed in 15 patients during 2020-2021.All patients underwent the operation,adjuvant radiotherapy,and hyperbaric oxygen therapy.All flap lengths and widths were recorded,and the blood perfusion of the flaps was measured on the first-day post-operation and the day of stitch removal.Flap survival and the quality of flaps were evaluated on the day of stitch removal.All harvested data were analyzed using the R package.Results:1.A logistic regression model and a nomogram were developed using the keloid activity assessment scale(KAAS)and the postoperative complications as predictors.The model achieved an area under the curve(AUC)of 0.871 in the training set of chest keloids and 0.802 in the test set of non-chest keloids,demonstrating its high predictive utility.2.Three prediction models of keloid recurrence were established based on machine learning methods.KAAS,mean arterial pressure level,postoperative complications,and inflammatory cell proportion played an important role in these models.The decision tree model outperformed the random forest and the logistic regression model in terms of accuracy,and the decision tree model was the most precise overall as well.In terms of AUC,logistic regression was the best-performing model,followed by random forest and decision trees.3.A total of 123 differentially expressed genes were found,including 45 up-regulated genes and 78 down-regulated genes.The expression level of CCR7 in the severe group was significantly higher than that in the mild group(P<0.05).The ROC curve showed that the expression level of CCR7 could distinguish the severe and mild groups,and the AUC was 0.833.The biological process categories enriched in GO included the regulation of T-cell activation and the positive regulation of leukocyte activation.The molecular functions mainly included cytokine receptor binding,amide binding,and immune receptor activation.Cell component analysis suggested that the cells were mainly located on the external side of plasma membrane and endocytic vesicle.KEGG pathway enrichment analysis was most enriched in the cytokine-cytokine receptor interaction signaling pathway,followed by the cell adhesion molecule pathway.4.The mean blood perfusion on the first-day post-operation(podl)and the day of stitch removal was 120.4013 and 168.6900,respectively(p=0.02249);2 flaps had partial necrosis(5.714%).Receiver operating characteristic(ROC)curve analysis showed that when the length/width ratio was less than 1.05,the quality of the flap was good(AUC=0.724),which suggests that the effective safe length/width ratio was 1.05.Conclusions:1.The KAAS can be used to predict recurrence.In this study,a nomogram was established to predict the recurrence of chest keloids within 2 years after surgery and adjuvant radiotherapy.2.Three prediction models of keloid recurrence were established based on machine learning methods,and it was found that the KAAS,blood pressure level,postoperative complications,and the proportion of inflammatory cells were involved.The three models were compared in different dimensions,the logistic model has the best prognostic effects in the aspect of AUC.3.The expression level of CCR7 indicates the severity of keloid,and regulating this gene may help to prevent or alleviate keloid or improve the effect of keloid treatment.The severity of keloid is closely related to immunomodulatory pathways.4.KSVNF is an applicable method for covering the remaining wound after keloid mass removal with sufficient blood perfusion and adequate skin quality.We recommend that the length/width ratio of the flap design not exceed 1.Combined with hyperbaric oxygen therapy can improve the blood supply of the flap and reduce the occurrence of postoperative complications.