| Due to the multifaceted heterogeneity of cancer,only some patients can benefit from drug therapy,and personalized drug use is important to improve the drug response rate and prognosis of cancer patients.It has been shown that transcriptomic data can reflect the key biological variants in cancer,and can also characterize the perturbation patterns of drugs on the organism.Bioinformatics studies based on transcriptomic data have made a series of achievements in the field of anti-cancer drug discovery and repurposing.However,the related studies cannot directly suggest personalized drug regimens for cancer patients,and there is a significant lack of research on the mechanism of action of drugs for specific cancer species.To this end,we combine drug computational methods and multi-source transcriptomic data to carry out the following work.Firstly,we construct a personalized drug discovery analysis system CPDR(Cancer Personalized Drug Recommendation)based on the reversal of disease signatures.CPDR has three features.1)It identifies the individual disease signature by using the patient subgroup with transcriptomic profiles similar to those of the input patient.2)Transcriptomic profile purification is supported for the subgroup with high infiltration of non-cancerous cells.3)It supports in silico drug efficacy assessment using drug sensitivity data on cancer cell lines.On the clinical pancreatic cancer dataset,the system can significantly distinguish the sensitive and resistant groups of gemcitabine treatment,and the area under the receiver operating characteristic curve(AUROC)can reach 0.77,showing the good performance of personalized drug prediction.We further use R language to build the above analysis system into a user-friendly package(Cancer Personalized Drug Recommendation,CPDR)to provide an analysis pipeline for cancer precision therapy research.Secondly,We applied CPDR to a personalized drug discovery study in colorectal cancer.We identified 10 drug candidates for five colorectal cancer patients,nine of which are in clinical trials or already on the market,and all of which can be validated in experimental cell line data.Notably,the histone deacetylase inhibitor romedipine was found to be a potential therapeutic agent in common for the five patients,which is consistent with the findings of the existing studies.Further studies revealed that PT1 patients were significantly sensitive to topoisomerase inhibitors,of which doxorubicin and valsoposide may be effective therapeutic agents for this patient.These results suggest that CPDR can identify known effective drugs for cancer,and also identify novel associations between individual cancer patients and drugs,providing targets and clues for basic research and clinical treatment.Finally,We applied CPDR to five GI cancers(including bile duct cancer,colorectal cancer,liver cancer,gastric cancer,and esophageal cancer)to explore drug intervention genes and pathways shared and differed between tumors.A total of 108 reversal genes for 1,517 samples were obtained,such as UBE2 C,CDK1,PLK1,BIRC5,and TOP2 A,which were significantly dysregulated in the cancer state but were significantly reversed after drug action,thus were valuable for the discovery of potential biomarkers and therapeutic targets.The analysis results suggest that TOP2 A can affect the development of GI cancer by participating in different pathways and targeting TOP2A-related pathways is a potential therapeutic strategy for this type of cancer,providing a new perspective and clues for the studies of the mechanism of drug action in GI cancer.In conclusion,CPDR is a promising analytical system for discovering personalized drug use for cancer patients and can be used to explore drug intervention genes and pathways,helping researchers to discover novel associations between cancer and drugs,as well as suggesting mechanisms of drug action in different cancer contexts. |
PDF Full Text Request