At present, molecular imaging and imaging omics research focuses on training artificial intelligence algorithms through clinical big data to predict IDH1 mutation status[6, 7], Ki-67 expression level and WHO classification of brain gliomas, in order to judge the prognosis of patients. However, pathological diagnosis, as the gold standard, it is not entirely accurate. In our clinical work, we have found that the prognosis of some patients is not as good as predicted by pathological diagnosis, because the area with the highest malignant degree of glioma is often unclear. At this stage, intraoperative imaging of glioma only shows the tumor boundary and does not indicate the core area of the tumor. Whether the tumor core is removed by surgery and whether the tumor core is detected by pathology are both problems.
We believe that the future research will focus on the fusion of imaging, clinical and pathological aspects. By utilizing preoperative multimodal magnetic resonance imaging, especially through molecular markers[8, 9], we can enhance the specificity of tumor imaging, identify the boundaries between gliomas and normal brain tissue, as well as the tumor core area, which is the most malignant part. Combined with surgical robot stereotactic orientation and intraoperative tumor imaging[10–12], we can delineate a reasonable surgical resection range and detect the tumor core with pathological specificity. we can provied feedback on the results, continuously correct them, improve the surgical resection rate and accuracy of pathological diagnosis, and optimize clinical treatment strategies.
Many scholars have used bioinformatics methods to screen prognostic related LncRNAs, with a total of hundreds, most of which only stay at the theoretical level and lack clinical practice. However, the process of tumor occurrence and development is extremely complex, and there are interactions between various genes, malcing it difficult for us to conduct clinical verification one by one. The same LncRNAs screened in different studies may suggest that they have a high possibility of playing a role in prognosis. This method of big data analysis to explore the inherent laws of diseases is precisely the connotation of bioinformatics. However, many of them are purely theoretical results, lacking experimental support, and further in-depth research is need to explore their internal mechanisms of action, combined with clinical practice, to find breakthroughs in possible targets.
Using bioinformatics methods and public databases, analyze the differential expression of LncRNAs between gliomas and normal brain tissue, as well as between low-grade and high-grade gliomas, calculate risk scores and establish prognostic models. Different case samples and different screening conditions can lead to different results. There is no absolutely accurate prognostic model (AUC cannot be 1), and the difference between the training set and the validation set will always exist. The accuracy of the prognostic model is not directly related to the number of LncRNAs it contains, and the overfitting of the algorithm leads to poor predictive performance of the model in unknown tests [13]. The frequency of the occurrence of LncRNAs does not necessarily indicate their importance in gliomas. How to screen the appropriate LncRNAs, design better algorithms, and comprehensively establish accurate and reliable prognostic models are the directions we will strive towards in the future. With the development of bioinformatics, including big data and artificial intelligence, we may not be able to cure tumors in the end, but we will better coexist with tumors.
LOC339529, also known as LINC02774, is a nuclear LncRNA. We and our collaborating unit found that it is downregulated in gliomas and decreases with decreasing malignancy, and its expression is negatively correlated with the relative index of enhanced magnetic resonance (RIEMR). Among the cases collected in our hospital, there were 2 cases of WHO grade 3 glioma and 1 case of WHO grade 4 glioma in the low expression group of LOC339529, and 3 cases of glioblastoma in the WHO high-grade group with high expression were all grade 4. The cooperative unit collected 56 samples, with 9 cases separately in the high expression group at grade 3 and 4, 6 cases in the low expression group at grade 3, and 18 cases in the low expression at grade 4, but no specific cell types were reported. Their cases have differences in glioma WHO grading and RIEMR, while ours only have differences in age. They found that high expression of LOC339529 in gliomas was associated with longer overall survival and better relapse-free survival in patients.
Although we did not find a correlation between high expression of LOC339529 and good prognosis in patients, and there is even a trend of low expression group being associated with good prognosis, this situation is not the first to be discovered. Glioblastoma has four molecular subgroups, namely, classical subgroup, mesenchymal subgroup, neural subgroup and proneural subgroup. U87-MG cells belong to mesenchymal type, and A172 cells belong to proneural type. CYTOR has been reported to have carcinogenic function in U87-MG cells, but acts as tumor suppressor in A172 glioblastoma cells [14]. Knocking down obvious therapeutic lncRNA targets in highly heterogeneous tumors such as glioblastoma may produce unnecessary tumor stimulation in some tumor cells, which suggests that we should further refine our reserch on tumors, as the same LncRNA may play different or even opposite roles in different subtypes of the same tumor.
This study is a prospective study with a small sample size. Due to the limited sample size and follow-up time, sufficient samples of various types of gliomas could not be collected. During the follow-up time of at least 18 months, clinical endpoints such as recurrence, progression or death could not be observed in most patients. Despite using strict experimental methods and repeating the experiment three times for each sample, we can not guarantee the impact of wax block production and storage on the detection of LOC339529 expression level. The primer sequences used by the cooperating unit was different from ours, and data integration with them was not carried out to reanalyze their impact on the prognosis of glioma patients. In highly heterogeneous tumors, the same LncRNA may play different or even opposite roles in different subtypes of the same tumor. Taking into account the various reasons mentioned above, we didnot find any correlation between high expression of LOC339529 and good prognosis in patients, but found there was even a trend that low expression groups being associated with good prognosis.
In this study, limited clinical data could not currently verify the correlation between high expression of LOC339529 and good prognosis in patients, and WHO high grade and Ki−67 index greater than or equal to 5% were associated with poor prognosis of glioma patients.