Abstract Background Anti-TNF agents were the only biologic for Inflammatory Bowel Disease (IBD) patients. However, new drugs targeting α4β7 integrin, IL-12/IL-23, and janus kinases were added to the IBD therapeutic armamentarium. Anti-TNFs are first-line biological agents in our country, and for that reason, the discrimination of patients who will be refractory of those therapies before treatment is relevant to prevent IBD patients from the exposition to multiple therapeutic failures. Therefore, we wanted to isolate robust and reproducible transcriptional markers able to specifically predict the response to anti-TNF therapy, which will contribute to targeted prevention, improve individual outcomes and decrease therapy expenditures. Methods We analysed publicly available microarray and RNAseq datasets of colon biopsies (GSE12251, GSE16879 and GSE73661) and whole blood cells from different cohorts of patients with Ulcerative Colitis (UC) (GSE191328). R package was used to perform the differential expression analyses and calculation of receiver operating characteristic (ROC). Cytoscape software and Morpheus (broadinstitute.org) was used to perform network data integration and visualization. Also, we used our local cohort of UC patients to further validate our results by RT-qPCR (n=21). Results We isolated a transcriptional signature composed of eight metalloproteases (MMPs) with high association with disease activity. Most of the transcripts of that signature were significantly upregulated in colon biopsies of non-responder patients to anti-TNFs before treatment in three different microarray datasets, but not in vedolizumab-treated patients. ROC curves of the different MMPs showed variability in the correlation to anti-TNF response among the different datasets analysed. Only three out of eight MMPs were still showing robust predictive power with an area under the curve around of 80% when all dataset were combined. Interestingly, PBMCs isolated from our local cohort of IBD patients presented increased expression of those MMPs in refractory patients to anti-TNF therapy, confirming that these three MMPs have the highest predictive power both in colon tissue and in peripheral mononuclear cells. Conclusion Overall, our study revealed a novel and robust MMP-transcriptional signature associated with anti-TNF but not vedolizumab therapy response. Nevertheless, our results should be corroborated in independent cohorts of patients with UC to test their efficacy and sensitivity in the clinical setting. Our data definitively contribute to the change from a reactive approach to predictive, preventive and personalised medicine.