Deep learning predicts patients outcome and mutations from digitized histology slides in gastrointestinal stromal tumor.

Risk assessment of gastrointestinal stromal tumor (GIST) according to the AFIP/Miettinen classification and mutational profiling are major tools for patient management. However, the AFIP/Miettinen classification depends heavily on mitotic counts, which is laborious and sometimes inconsistent between pathologists. It has also been shown to be imperfect in stratifying patients. Molecular testing is costly and time-consuming, therefore, not systematically performed in all countries. New methods to improve risk and molecular predictions are hence crucial to improve the tailoring of adjuvant therapy. We have built deep learning (DL) models on digitized HES-stained whole slide images (WSI) to predict patients' outcome and mutations. Models were trained with a cohort of 1233 GIST and validated on an independent cohort of 286 GIST. DL models yielded comparable results to the Miettinen classification for relapse-free-survival prediction in localized GIST without adjuvant Imatinib (C-index=0.83 in cross-validation and 0.72 for independent testing). DL splitted Miettinen intermediate risk GIST into high/low-risk groups (p value = 0.002 in the training set and p value = 0.29 in the testing set). DL models achieved an area under the receiver operating characteristic curve (AUC) of 0.81, 0.91, and 0.71 for predicting mutations in KIT, PDGFRA and wild type, respectively, in cross-validation and 0.76, 0.90, and 0.55 in independent testing. Notably, PDGFRA exon18 D842V mutation, which is resistant to Imatinib, was predicted with an AUC of 0.87 and 0.90 in cross-validation and independent testing, respectively. Additionally, novel histological criteria predictive of patients' outcome and mutations were identified by reviewing the tiles selected by the models. As a proof of concept, our study showed the possibility of implementing DL with digitized WSI and may represent a reproducible way to improve tailoring therapy and precision medicine for patients with GIST.

Inactivating mutations in genes encoding for components of the BAF/PBAF complex and immune-checkpoint inhibitor outcome.

Alterations of genes encoding subunits of the BAF/PBAF complexes are among the most frequent gene aberrations in human cancer. Such alterations have been shown to have an impact on tumor microenvironnement and on the capacity of tumors to respond to immune-checkpoint inhibitors (ICI). We analysed the clinical and genetic data from 43,728 patients accessed through cBioportal. The mutational frequencies of ARID1A, ARID1B, ARID2, PBRM1, SMARCA4, and SMARCB1 were 6.6%, 3,4, 3.4, 3.2, 4.1, and 1.2%, respectively. We then investigated the association between the presence of least one nonsynonymous somatic mutation of ARID1A, ARID1B, ARID2, PBRM1, SMARCA4, or SMARCB1 and overall survival of 1661 patients treated with an ICI. Across the entire cohort, patients with BAF/PBAF mutated tumors have a statistically significant improvement in overall survival (median overall survival: 28 months [95% CI 21.6-34.3] versus 15 months [95% CI 12.9-17.0], p < 0.0001). When tumor mutational burden was adjusted for a multivariable Cox regression analysis, BAF/PBAF gene mutations remained an independent prognostic factor for overall survival in patients treated ICI. Our results establish a relationship between mutations in key genes encoding for components of the BAF/PBAF complex and outcome of patients treated with ICI. Further studies are needed to elucidate the underlying mechanisms of this interaction.

Integration of miRNA-regulatory networks in hepatic stellate cells identifies TIMP3 as a key factor in chronic liver disease.

BACKGROUND & AIMS: Activation of hepatic stellate cells (HSC) is a critical process involved in liver fibrosis. Several miRNAs are implicated in gene regulation during this process but their exact and respective contribution is still incompletely understood. Here we propose an integrative approach of miRNA-regulatory networks to predict new targets. METHODS: miRNA regulatory networks in activated HSCs were built using lists of validated miRNAs and the CyTargetLinker tool. The resulting graphs were filtered according to public transcriptomic data and the reduced graphs were analysed through GO annotation. A miRNA network regulating the expression of TIMP3 was further studied in human liver samples, isolated hepatic cells and mouse model of liver fibrosis. RESULTS: Within the up-regulated miRNAs, we identified a subnetwork of five miRNAs (miR-21-5p, miR-222-3p, miR-221-3p miR-181b-5p and miR-17-5p) that target TIMP3. We demonstrated that TIMP3 expression is inversely associated with inflammatory activity and IL1-ß expression in vivo. We further showed that IL1-ß inhibits TIMP3 expression in HSC-derived LX-2 cells. Using data from The Cancer Genome Atlas (TCGA), we showed that, in hepatocellular carcinoma (HCC), TIMP3 expression is associated with survival (P < .001), while miR-221 (P < .05), miR-222 (P < .01) and miR-181b (P < .01) are markers for a poor prognosis. CONCLUSIONS: Several miRNAs targeting TIMP3 are up-regulated in activated HSCs and down-regulation of TIMP3 expression is associated with inflammatory activity in liver fibrosis and poor prognosis in HCC. The regulatory network including specific miRNAs and TIMP3 is therefore central for the evolution of chronic liver disease.