Highlights d Ultra-deep rRNA-depleted RNA sequencing of 144 localized prostate tumors d Fusion gene profiles differentiate localized from metastatic disease d Widespread RNA circularization events define clinically distinct tumor subtypes d Functional screening reveals pervasive circular isoformspecific essentiality

Summary Background Improvements to prognostic models in metastatic castration-resistant prostate cancer have the potential to augment clinical trial design and guide treatment strategies. In partnership with Project Data Sphere, a not-for-profit initiative allowing data from cancer clinical trials to be shared broadly with researchers, we designed an open-data, crowdsourced, DREAM (Dialogue for Reverse Engineering Assessments and Methods) challenge to not only identify a better prognostic model for prediction of survival in patients with metastatic castration-resistant prostate cancer but also engage a community of international data scientists to study this disease. Methods Data from the comparator arms of four phase 3 clinical trials in first-line metastatic castration-resistant prostate cancer were obtained from Project Data Sphere, comprising 476 patients treated with docetaxel and prednisone from the ASCENT2 trial, 526 patients treated with docetaxel, prednisone, and placebo in the MAINSAIL trial, 598 patients treated with docetaxel, prednisone or prednisolone, and placebo in the VENICE trial, and 470 patients treated with docetaxel and placebo in the ENTHUSE 33 trial. Datasets consisting of more than 150 clinical variables were curated centrally, including demographics, laboratory values, medical history, lesion sites, and previous treatments. Data from ASCENT2, MAINSAIL, and VENICE were released publicly to be used as training data to predict the outcome of interest—namely, overall survival. Clinical data were also released for ENTHUSE 33, but data for outcome variables (overall survival and event status) were hidden from the challenge participants so that ENTHUSE 33 could be used for independent validation. Methods were evaluated using the integrated time-dependent area under the curve (iAUC). The reference model, based on eight clinical variables and a penalised Cox proportional-hazards model, was used to compare method performance. Further validation was done using data from a fifth trial—ENTHUSE M1—in which 266 patients with metastatic castration-resistant prostate cancer were treated with placebo alone. Findings 50 independent methods were developed to predict overall survival and were evaluated through the DREAM challenge. The top performer was based on an ensemble of penalised Cox regression models (ePCR), which uniquely identified predictive interaction effects with immune biomarkers and markers of hepatic and renal function. Overall, ePCR outperformed all other methods (iAUC 0·791; Bayes factor >5) and surpassed the reference model (iAUC 0·743; Bayes factor >20). Both the ePCR model and reference models stratified patients in the ENTHUSE 33 trial into high-risk and low-risk groups with significantly different overall survival (ePCR: hazard ratio 3·32, 95% CI 2·39–4·62, p<0·0001; reference model: 2·56, 1·85–3·53, p<0·0001). The new model was validated further on the ENTHUSE M1 cohort with similarly high performance (iAUC 0·768). Meta-analysis across all methods confirmed previously identified...

The intrinsically disordered regions of eukaryotic proteomes are enriched in short linear motifs (SLiMs), which are of crucial relevance for cellular signaling and protein regulation; many mediate interactions by providing binding sites for peptide-binding domains. The vast majority of SLiMs remain to be discovered highlighting the need for experimental methods for their large-scale identification. We present a novel proteomic peptide phage display (ProP-PD) library that displays peptides representing the disordered regions of the human proteome, allowing direct large-scale interrogation of most potential binding SLiMs in the proteome. The performance of the ProP-PD library was validated through selections against SLiM-binding bait domains with distinct folds and binding preferences. The vast majority of identified binding peptides contained sequences that matched the known SLiM-binding specificities of the bait proteins. For SHANK1 PDZ, we establish a novel consensus TxF motif for its non-C-terminal ligands. The binding peptides mostly represented novel target proteins, however, several previously validated protein-protein interactions (PPIs) were also discovered. We determined the affinities between the VHS domain of GGA1 and three identified ligands to 40-130 lM through isothermal titration calorimetry, and confirmed interactions through coimmunoprecipitation using fulllength proteins. Taken together, we outline a general pipeline for the design and construction of ProP-PD libraries and the analysis of ProP-PDderived, SLiM-based PPIs. We demonstrated the methods potential to identify low affinity motif-mediated interactions for modular domains with distinct binding preferences. The approach is a highly useful complement to the current toolbox of methods for PPI discovery.

Biomarkers are rapidly gaining importance in personalized medicine. Although numerous molecular signatures have been developed over the past decade, there is a lack of overlap and many biomarkers fail to validate in independent patient cohorts and hence are not useful for clinical application. For these reasons, identification of novel and robust biomarkers remains a formidable challenge. We combine targeted proteomics with computational biology to discover robust proteomic signatures for prostate cancer. Quantitative proteomics conducted in expressed prostatic secretions from men with extraprostatic and organ-confined prostate cancers identified 133 differentially expressed proteins. Using synthetic peptides, we evaluate them by targeted proteomics in a 74-patient cohort of expressed prostatic secretions in urine. We quantify a panel of 34 candidates in an independent 207-patient cohort. We apply machine-learning approaches to develop clinical predictive models for prostate cancer diagnosis and prognosis. Our results demonstrate that computationally guided proteomics can discover highly accurate non-invasive biomarkers.

Protein-protein interactions (PPIs) are emerging as a promising new class of drug targets. Here, we present a novel high-throughput approach to screen inhibitors of PPIs in cells. We designed a library of 50,000 human peptide binding motifs and used a pooled lentiviral system to express them intracellularly and screen for their effects on cell proliferation. We thereby identified inhibitors that drastically reduced the viability of a pancreas cancer line (RWP1) while leaving a control line virtually unaffected. We identified their target interactions computationally, and validated a subset in experiments. We also discovered their potential mechanisms of action including apoptosis and cell cycle arrest. Finally, we confirmed that synthetic lipopeptide versions of our inhibitors have similarly specific and dosage dependent effects on cancer cell growth. Our screen reveals new drug targets and peptide drug leads and it provides a rich dataset covering phenotypes for inhibition of thousands of interactions.