Michael J. Mason

To develop machine learning methods to quantify joint damage in patients with rheumatoid arthritis (RA), we developed the RA2 DREAM Challenge, a crowdsourced competition that utilized existing radiographic images and “gold-standard” scores on 674 sets of films from 562 patients. Training and leaderboard sets were provided to participants to develop methods to quantify joint space narrowing and erosions. In the final round, participants submitted containerized codes on a test set; algorithms were evaluated using weighted root mean square error (RMSE). In the leaderboard round, there were 173 submissions from 26 teams in 7 countries. Of the 13 submissions in the final round, four top-performing teams were identified. Robustness of results was assessed using Bayes factor and validated using an independent set of radiographs. The top-performing algorithms, which consisted of different styles of deep learning models, provided accurate and robust quantification of joint damage in RA. Ultimately, these methods lay the groundwork to accelerate research and help clinicians to optimize treatments to minimize joint damage.

Despite decades of intensive search for compounds that modulate the activity of particular protein targets, a large proportion of the human kinome remains as yet undrugged. Effective approaches are therefore required to map the massive space of unexplored compound–kinase interactions for novel and potent activities. Here, we carry out a crowdsourced benchmarking of predictive algorithms for kinase inhibitor potencies across multiple kinase families tested on unpublished bioactivity data. We find the top-performing predictions are based on various models, including kernel learning, gradient boosting and deep learning, and their ensemble leads to a predictive accuracy exceeding that of single-dose kinase activity assays. We design experiments based on the model predictions and identify unexpected activities even for under-studied kinases, thereby accelerating experimental mapping efforts. The open-source prediction algorithms together with the bioactivities between 95 compounds and 295 kinases provide a resource for benchmarking prediction algorithms and for extending the druggable kinome.

Despite recent advancements in the treatment of multiple myeloma (MM), nearly all patients ultimately relapse and many become refractory to their previous therapies. Although many therapies exist with diverse mechanisms of action, it is not yet clear how the differences in MM biology across patients impacts the likelihood of success for existing therapies and those in the pipeline. Therefore, we not only need the ability to predict which patients are at high risk for disease progression, but also a means to understand the mechanisms underlying their risk. We hypothesized that knowledge of the biological networks that give rise to MM, specifically the transcriptional regulatory network (TRN) and the mechanisms by which mutations impact gene regulation, would enable improved predictions of disease progression and actionable insights for treatment. Here we present a method to infer TRNs from multi-omics data and apply it to the generation of a MM TRN that links chromosomal abnormalities and somatic mutations to downstream effects on gene expression via perturbation of transcriptional regulators. We find that 141 genetic programs underlie the disease and that the activity profile of these programs fall into one of 25 distinct transcriptional states. These transcriptional signatures prove to be more predictive of outcomes than do mutations and reveal plausible mechanisms for relapse, including the establishment of an immuno-suppressive microenvironment. Moreover, we observe subtype-specific vulnerabilities to interventions with existing drugs and motivate the development of new targeted therapies that appear especially promising for relapsed refractory MM.

While the past decade has seen meaningful improvements in clinical outcomes for multiple myeloma patients, a subset of patients does not benefit from current therapeutics for unclear reasons. Many gene expression-based models of risk have been developed, but each model uses a different combination of genes and often involves assaying many genes making them difficult to implement. We organized the Multiple Myeloma DREAM Challenge, a crowdsourced effort to develop models of rapid progression in newly diagnosed myeloma patients and to benchmark these against previously published models. This effort lead to more robust predictors and found that incorporating specific demographic and clinical features improved gene expression-based models of high risk. Furthermore, post-challenge analysis identified a novel expression-based risk marker, PHF19, which has recently been found to have an important biological role in multiple myeloma. Lastly, we show that a simple four feature predictor composed of age, ISS, and expression of PHF19 and MMSET performs similarly to more complex models with many more gene expression features included.

Despite decades of intensive search for compounds that modulate the activity of particular targets, there are currently small-molecules available only for a small proportion of the human proteome. Effective approaches are therefore required to map the massive space of unexplored compound-target interactions for novel and potent activities. Here, we carried out a crowdsourced benchmarking of predictive models for kinase inhibitor potencies across multiple kinase families using unpublished bioactivity data. The top-performing predictions were based on kernel learning, gradient boosting and deep learning, and their ensemble resulted in predictive accuracy exceeding that of kinase activity assays. We then made new experiments based on the model predictions, which further improved the accuracy of experimental mapping efforts and identified unexpected potencies even for under-studied kinases. The open-source algorithms together with the novel bioactivities between 95 compounds and 295 kinases provide a resource for benchmarking new prediction algorithms and for extending the druggable kinome.

It remains unknown to what extent consensus molecular subtype (CMS) groups and immune-stromal infiltration patterns improve our ability to predict outcomes over tumor-node-metastasis (TNM) staging and microsatellite instability (MSI) status in early-stage colorectal cancer (CRC). We carried out a comprehensive retrospective biomarker analysis of prognostic markers in adjuvant chemotherapy-untreated (N = 1656) and treated (N = 980), stage II (N = 1799) and III (N = 837) CRCs. We defined CMS scores and estimated CD8+ cytotoxic lymphocytes (CytoLym) and cancer-associated fibroblasts (CAF) infiltration scores from bulk tumor tissue transcriptomes (CMSclassifier and MCPcounter R packages); constructed a stratified multivariable Cox model for disease-free survival (DFS); and calculated the relative proportion of explained variation by each marker (clinicopathological [ClinPath], genomics [Gen: MSI, BRAF and KRAS mutations], CMS scores [CMS] and microenvironment cells [MicroCells: CytoLym+CAF]). In multivariable models, only ClinPath and MicroCells remained significant prognostic factors, with both CytoLym and CAF infiltration scores improving survival prediction beyond other markers. The explained variation for DFS models of ClinPath, MicroCells, Gen markers and CMS4 scores was 77%, 14%, 5.3% and 3.7%, respectively, in stage II; and 55.9%, 35.1%, 4.1% and 0.9%, respectively, in stage III. Patients whose tumors were CytoLym high/CAF low had better DFS than other strata [HR=0.71 (0.6-0.9); P = 0.004]. Microsatellite stable tumors had the strongest signal for improved outcomes with CytoLym high scores (interaction P = 0.04) and the poor prognosis linked to high CAF scores was limited to stage III disease (interaction P = 0.04). Our results confirm that tumor microenvironment infiltration patterns represent potent determinants of the risk for distant dissemination in early-stage CRC. Multivariable models suggest that the prognostic value of MSI and CMS groups is largely explained by CytoLym and CAF infiltration patterns.

While the past decade has seen meaningful improvements in clinical outcomes for multiple myeloma patients, a subset of patients do not benefit from current therapeutics for unclear reasons. Many gene expression-based models of risk have been developed, but each model uses a different combination of genes and often involve assaying many genes making them difficult to implement. We organized the Multiple Myeloma DREAM Challenge, a crowdsourced effort to develop models of rapid progression in newly diagnosed myeloma patients and to benchmark these against previously published models. This effort lead to more robust predictors and found that incorporating specific demographic and clinical features improved gene expression-based models of high risk. Furthermore, post challenge analysis identified a novel expression-based risk marker and histone modifier, PHF19 , which featured prominently in several independent models. Lastly, we show that a simple four feature predictor composed of age, International Staging System stage (ISS), and expression of PHF19 and MMSET performs similarly to more complex models with many more gene expression features included. Key points: Most comprehensive and unbiased assessment of prognostic biomarkers in MM resulting in a robust and parsimonious model. Identification of PHF19 as the expression based biomarker most strongly associated with rapid progression in MM patients.

The effectiveness of most cancer targeted therapies is short-lived. Tumors often develop resistance that might be overcome with drug combinations. However, the number of possible combinations is vast, necessitating data-driven approaches to find optimal patient-specific treatments. Here we report AstraZeneca’s large drug combination dataset, consisting of 11,576 experiments from 910 combinations across 85 molecularly characterized cancer cell lines, and results of a DREAM Challenge to evaluate computational strategies for predicting synergistic drug pairs and biomarkers. 160 teams participated to provide a comprehensive methodological development and benchmarking. Winning methods incorporate prior knowledge of drug-target interactions. Synergy is predicted with an accuracy matching biological replicates for >60% of combinations. However, 20% of drug combinations are poorly predicted by all methods. Genomic rationale for synergy predictions are identified, including ADAM17 inhibitor antagonism when combined with PIK3CB/D inhibition contrasting to synergy when combined with other PI3K-pathway inhibitors in PIK3CA mutant cells.

Unmasking an agent of inflammatory anemia: Infectious and autoimmune diseases are associated with anemia and thrombocytopenia. A severe form of inflammatory cytopenia called macrophage activation syndrome (MAS) may occur during rheumatological disorders and viral infections. Akilesh et al. show that monocyte recognition of self- or pathogen-derived nucleic acids via Toll-like receptors 7 and 9 (TLR7 and TLR9) drives MAS-like disease in mice. TLR7 or TLR9 signaling in monocytes causes these cells to differentiate into inflammatory hematophagocytes (iHPCs), which are similar to but distinct from red pulp macrophages. Preventing iHPC differentiation by depleting monocytes relieves MAS-like symptoms. When mice were subjected to a model of malarial anemia, MyD88- and endosomal TLR-dependent iHPC differentiation also occurred. Thus, iHPCs may play a role in both MAS-driven and malarial anemia, as well as thrombocytopenia. Science , this issue p. eaao5213

Purpose: KRAS mutation is a common canonical mutation in colorectal cancer, found at differing frequencies in all consensus molecular subtypes (CMS). The independent immunobiological impacts of RAS mutation and CMS are unknown. Thus, we explored the immunobiological effects of KRAS mutation across the CMS spectrum. Experimental Design: Expression analysis of immune genes/signatures was performed using The Cancer Genome Atlas (TCGA) RNA-seq and the KFSYSCC microarray datasets. Multivariate analysis included KRAS status, CMS, tumor location, MSI status, and neoantigen load. Protein expression of STAT1, HLA-class II, and CXCL10 was analyzed by digital IHC. Results: The Th1-centric co-ordinate immune response cluster (CIRC) was significantly, albeit modestly, reduced in KRAS-mutant colorectal cancer in both datasets. Cytotoxic T cells, neutrophils, and the IFNγ pathway were suppressed in KRAS-mutant samples. The expressions of STAT1 and CXCL10 were reduced at the mRNA and protein levels. In multivariate analysis, KRAS mutation, CMS2, and CMS3 were independently predictive of reduced CIRC expression. Immune response was heterogeneous across KRAS-mutant colorectal cancer: KRAS-mutant CMS2 samples have the lowest CIRC expression, reduced expression of the IFNγ pathway, STAT1 and CXCL10, and reduced infiltration of cytotoxic cells and neutrophils relative to CMS1 and CMS4 and to KRAS wild-type CMS2 samples in the TCGA. These trends held in the KFSYSCC dataset. Conclusions: KRAS mutation is associated with suppressed Th1/cytotoxic immunity in colorectal cancer, the extent of the effect being modulated by CMS subtype. These results add a novel immunobiological dimension to the biological heterogeneity of colorectal cancer. Clin Cancer Res; 24(1); 224–33. ©2017 AACR.