Identifying important microbial and genomic biomarkers for differentiating right- versus left-sided colorectal cancer using random forest models.

BACKGROUND: Colorectal cancer (CRC) is a heterogeneous disease, with subtypes that have different clinical behaviours and subsequent prognoses. There is a growing body of evidence suggesting that right-sided colorectal cancer (RCC) and left-sided colorectal cancer (LCC) also differ in treatment success and patient outcomes. Biomarkers that differentiate between RCC and LCC are not well-established. Here, we apply random forest (RF) machine learning methods to identify genomic or microbial biomarkers that differentiate RCC and LCC. METHODS: RNA-seq expression data for 58,677 coding and non-coding human genes and count data for 28,557 human unmapped reads were obtained from 308 patient CRC tumour samples. We created three RF models for datasets of human genes-only, microbes-only, and genes-and-microbes combined. We used a permutation test to identify features of significant importance. Finally, we used differential expression (DE) and paired Wilcoxon-rank sum tests to associate features with a particular side. RESULTS: RF model accuracy scores were 90%, 70%, and 87% with area under curve (AUC) of 0.9, 0.76, and 0.89 for the human genomic, microbial, and combined feature sets, respectively. 15 features were identified as significant in the model of genes-only, 54 microbes in the model of microbes-only, and 28 genes and 18 microbes in the model with genes-and-microbes combined. PRAC1 expression was the most important feature for differentiating RCC and LCC in the genes-only model, with HOXB13, SPAG16, HOXC4, and RNLS also playing a role. Ruminococcus gnavus and Clostridium acetireducens were the most important in the microbial-only model. MYOM3, HOXC4, Coprococcus eutactus, PRAC1, lncRNA AC012531.25, Ruminococcus gnavus, RNLS, HOXC6, SPAG16 and Fusobacterium nucleatum were most important in the combined model. CONCLUSIONS: Many of the identified genes and microbes among all models have previously established associations with CRC. However, the ability of RF models to account for inter-feature relationships within the underlying decision trees may yield a more sensitive and biologically interconnected set of genomic and microbial biomarkers.

A p53 transcriptional signature in primary and metastatic cancers derived using machine learning.

The tumor suppressor gene, TP53, has the highest rate of mutation among all genes in human cancer. This transcription factor plays an essential role in the regulation of many cellular processes. Mutations in TP53 result in loss of wild-type p53 function in a dominant negative manner. Although TP53 is a well-studied gene, the transcriptome modifications caused by the mutations in this gene have not yet been explored in a pan-cancer study using both primary and metastatic samples. In this work, we used a random forest model to stratify tumor samples based on TP53 mutational status and detected a p53 transcriptional signature. We hypothesize that the existence of this transcriptional signature is due to the loss of wild-type p53 function and is universal across primary and metastatic tumors as well as different tumor types. Additionally, we showed that the algorithm successfully detected this signature in samples with apparent silent mutations that affect correct mRNA splicing. Furthermore, we observed that most of the highly ranked genes contributing to the classification extracted from the random forest have known associations with p53 within the literature. We suggest that other genes found in this list including GPSM2, OR4N2, CTSL2, SPERT, and RPE65 protein coding genes have yet undiscovered linkages to p53 function. Our analysis of time on different therapies also revealed that this signature is more effective than the recorded TP53 status in detecting patients who can benefit from platinum therapies and taxanes. Our findings delineate a p53 transcriptional signature, expand the knowledge of p53 biology and further identify genes important in p53 related pathways.

The Diverse Genomic Landscape of Clinically Low-risk Prostate Cancer.

BACKGROUND: Among men with clinically low-risk prostate cancer, we have previously documented heterogeneity in terms of clinical characteristics and genomic risk scores. OBJECTIVE: To further study the underlying tumor biology of this patient population, by interrogating broader patterns of gene expression among men with clinically low-risk tumors. DESIGN, SETTING, AND PARTICIPANTS: Prostate biopsies from 427 patients considered potentially suitable for active surveillance underwent central pathology review and genome-wide expression profiling. These cases were compared with 1290 higher-risk biopsy cases with diverse clinical features from a prospective genomic registry. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: Average genomic risk (AGR) was determined from 18 published prognostic signatures, and MSigDB hallmark gene sets were analyzed using bootstrapped clustering methods. These sets were examined in relation to clinical variables and pathological and biochemical outcomes using multivariable regression analysis. RESULTS AND LIMITATIONS: A total of 408 (96%) biopsies passed RNA quality control. Based on AGR quartiles defined by the high-risk multicenter cases, the University of California, San Francisco (UCSF) low-risk patients were distributed across the quartiles as 219 (54%), 107 (26%), 61 (15%), and 21 (5%). Unsupervised clustering analysis of the hallmark gene set scores revealed three clusters, which were enriched for the previously described PAM50 luminal A, luminal B, and basal subtypes. AGR, but not the clusters, was associated with both pathological (odds ratio 1.34, 95% confidence interval [CI] 1.14-1.58) and biochemical outcomes (hazard ratio 1.53, 95% CI 1.19-1.93). These results may underestimate within-prostate genomic heterogeneity. CONCLUSIONS: Prostate cancers that are homogeneously low risk by traditional characteristics demonstrate substantial diversity at the level of genomic expression. Molecular substratification of low-risk prostate cancer will yield a better understanding of its divergent biology and, in the future may help personalize treatment recommendations. PATIENT SUMMARY: We studied the genomic characteristics of tumors from men diagnosed with low-risk prostate cancer. We found three main subtypes of prostate cancer with divergent tumor biology, similar to what has previously been found in women with breast cancer. In addition, we found that genomic risk scores were associated with worse pathology findings and prostate-specific antigen recurrence after surgery. These results suggest even greater genomic diversity among low-risk patients than has previously been documented with more limited signatures.

Development and Validation of a Novel Integrated Clinical-Genomic Risk Group Classification for Localized Prostate Cancer.

Purpose It is clinically challenging to integrate genomic-classifier results that report a numeric risk of recurrence into treatment recommendations for localized prostate cancer, which are founded in the framework of risk groups. We aimed to develop a novel clinical-genomic risk grouping system that can readily be incorporated into treatment guidelines for localized prostate cancer. Materials and Methods Two multicenter cohorts (n = 991) were used for training and validation of the clinical-genomic risk groups, and two additional cohorts (n = 5,937) were used for reclassification analyses. Competing risks analysis was used to estimate the risk of distant metastasis. Time-dependent c-indices were constructed to compare clinicopathologic risk models with the clinical-genomic risk groups. Results With a median follow-up of 8 years for patients in the training cohort, 10-year distant metastasis rates for National Comprehensive Cancer Network (NCCN) low, favorable-intermediate, unfavorable-intermediate, and high-risk were 7.3%, 9.2%, 38.0%, and 39.5%, respectively. In contrast, the three-tier clinical-genomic risk groups had 10-year distant metastasis rates of 3.5%, 29.4%, and 54.6%, for low-, intermediate-, and high-risk, respectively, which were consistent in the validation cohort (0%, 25.9%, and 55.2%, respectively). C-indices for the clinical-genomic risk grouping system (0.84; 95% CI, 0.61 to 0.93) were improved over NCCN (0.73; 95% CI, 0.60 to 0.86) and Cancer of the Prostate Risk Assessment (0.74; 95% CI, 0.65 to 0.84), and 30% of patients using NCCN low/intermediate/high would be reclassified by the new three-tier system and 67% of patients would be reclassified from NCCN six-tier (very-low- to very-high-risk) by the new six-tier system. Conclusion A commercially available genomic classifier in combination with standard clinicopathologic variables can generate a simple-to-use clinical-genomic risk grouping that more accurately identifies patients at low, intermediate, and high risk for metastasis and can be easily incorporated into current guidelines to better risk-stratify patients.

Low PCA3 expression is a marker of poor differentiation in localized prostate tumors: exploratory analysis from 12,076 patients.

BACKGROUND: Prostate cancer antigen 3 (PCA3) is a prostate cancer diagnostic biomarker that has been clinically validated. The limitations of the diagnostic role of PCA3 in initial biopsy and the prognostic role are not well established. Here, we elucidate the limitations of tissue PCA3 to predict high grade tumors in initial biopsy. RESULTS: PCA3 has a bimodal distribution in both biopsy and radical prostatectomy (RP) tissues, where low PCA3 expression was significantly associated with high grade disease (p<0.001). PCA3 had a poor performance of predicting high grade disease in initial biopsy (GS≥8) with 55% sensitivity and high false negative rates; 42% of high Gleason (≥8) samples had low PCA3. In RP, low PCA3 is associated with adverse pathological features, clinical recurrence outcome and greater probability of metastatic progression (p<0.001). MATERIALS AND METHODS: A total of 1,694 expression profiles from biopsy and 10,382 from RP patients with high risk tumors were obtained from the Decipher Genomic Resource Information Database (GRIDTM)prostate cancer database. The primary clinical endpoint was distant metastasis-free survival for RP and high Gleason grade for biopsy. Logistic regression analyses and Cox proportional hazards models were used to evaluate the association of PCA3 with clinical variables and risk of metastasis. CONCLUSIONS: There is high prevalence of high grade tumors with low PCA3 expression in the biopsy setting. Therefore, urologists should be warned that using PCA3 as stand-alone test may lead to high rate of under-diagnosis of high grade disease in initial biopsy setting.

Ability of a Genomic Classifier to Predict Metastasis and Prostate Cancer-specific Mortality after Radiation or Surgery based on Needle Biopsy Specimens.

BACKGROUND: Decipher is a validated genomic classifier developed to determine the biological potential for metastasis after radical prostatectomy (RP). OBJECTIVE: To evaluate the ability of biopsy Decipher to predict metastasis and Prostate cancer-specific mortality (PCSM) in primarily intermediate- to high-risk patients treated with RP or radiation therapy (RT). DESIGN, SETTING, AND PARTICIPANTS: Two hundred and thirty-five patients treated with either RP (n=105) or RT±androgen deprivation therapy (n=130) with available genomic expression profiles generated from diagnostic biopsy specimens from seven tertiary referral centers. The highest-grade core was sampled and Decipher was calculated based on a locked random forest model. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: Metastasis and PCSM were the primary and secondary outcomes of the study, respectively. Cox analysis and c-index were used to evaluate the performance of Decipher. RESULTS AND LIMITATIONS: With a median follow-up of 6 yr among censored patients, 34 patients developed metastases and 11 died of prostate cancer. On multivariable analysis, biopsy Decipher remained a significant predictor of metastasis (hazard ratio: 1.37 per 10% increase in score, 95% confidence interval [CI]: 1.06-1.78, p=0.018) after adjusting for clinical variables. For predicting metastasis 5-yr post-biopsy, Cancer of the Prostate Risk Assessment score had a c-index of 0.60 (95% CI: 0.50-0.69), while Cancer of the Prostate Risk Assessment plus biopsy Decipher had a c-index of 0.71 (95% CI: 0.60-0.82). National Comprehensive Cancer Network risk group had a c-index of 0.66 (95% CI: 0.53-0.77), while National Comprehensive Cancer Network plus biopsy Decipher had a c-index of 0.74 (95% CI: 0.66-0.82). Biopsy Decipher was a significant predictor of PCSM (hazard ratio: 1.57 per 10% increase in score, 95% CI: 1.03-2.48, p=0.037), with a 5-yr PCSM rate of 0%, 0%, and 9.4% for Decipher low, intermediate, and high, respectively. CONCLUSIONS: Biopsy Decipher predicted metastasis and PCSM from diagnostic biopsy specimens of primarily intermediate- and high-risk men treated with first-line RT or RP. PATIENT SUMMARY: Biopsy Decipher predicted metastasis and prostate cancer-specific mortality risk from diagnostic biopsy specimens.

Reusable, extensible, and modifiable R scripts and Kepler workflows for comprehensive single set ChIP-seq analysis.

BACKGROUND: There has been an enormous expansion of use of chromatin immunoprecipitation followed by sequencing (ChIP-seq) technologies. Analysis of large-scale ChIP-seq datasets involves a complex series of steps and production of several specialized graphical outputs. A number of systems have emphasized custom development of ChIP-seq pipelines. These systems are primarily based on custom programming of a single, complex pipeline or supply libraries of modules and do not produce the full range of outputs commonly produced for ChIP-seq datasets. It is desirable to have more comprehensive pipelines, in particular ones addressing common metadata tasks, such as pathway analysis, and pipelines producing standard complex graphical outputs. It is advantageous if these are highly modular systems, available as both turnkey pipelines and individual modules, that are easily comprehensible, modifiable and extensible to allow rapid alteration in response to new analysis developments in this growing area. Furthermore, it is advantageous if these pipelines allow data provenance tracking. RESULTS: We present a set of 20 ChIP-seq analysis software modules implemented in the Kepler workflow system; most (18/20) were also implemented as standalone, fully functional R scripts. The set consists of four full turnkey pipelines and 16 component modules. The turnkey pipelines in Kepler allow data provenance tracking. Implementation emphasized use of common R packages and widely-used external tools (e.g., MACS for peak finding), along with custom programming. This software presents comprehensive solutions and easily repurposed code blocks for ChIP-seq analysis and pipeline creation. Tasks include mapping raw reads, peakfinding via MACS, summary statistics, peak location statistics, summary plots centered on the transcription start site (TSS), gene ontology, pathway analysis, and de novo motif finding, among others. CONCLUSIONS: These pipelines range from those performing a single task to those performing full analyses of ChIP-seq data. The pipelines are supplied as both Kepler workflows, which allow data provenance tracking, and, in the majority of cases, as standalone R scripts. These pipelines are designed for ease of modification and repurposing.

An inhaled dose of budesonide induces genes involved in transcription and signaling in the human airways: enhancement of anti- and proinflammatory effector genes.

Although inhaled glucocorticoids, or corticosteroids (ICS), are generally effective in asthma, understanding their anti-inflammatory actions in vivo remains incomplete. To characterize glucocorticoid-induced modulation of gene expression in the human airways, we performed a randomized placebo-controlled crossover study in healthy male volunteers. Six hours after placebo or budesonide inhalation, whole blood, bronchial brushings, and endobronchial biopsies were collected. Microarray analysis of biopsy RNA, using stringent (≥2-fold, 5% false discovery rate) or less stringent (≥1.25-fold, P ≤ 0.05) criteria, identified 46 and 588 budesonide-induced genes, respectively. Approximately two third of these genes are transcriptional regulators (KLF9, PER1, TSC22D3, ZBTB16), receptors (CD163, CNR1, CXCR4, LIFR, TLR2), or signaling genes (DUSP1, NFKBIA, RGS1, RGS2, ZFP36). Listed genes were qPCR verified. Expression of anti-inflammatory and other potentially beneficial genes is therefore confirmed and consistent with gene ontology (GO) terms for negative regulation of transcription and gene expression. However, GO terms for transcription, signaling, metabolism, proliferation, inflammatory responses, and cell movement were also associated with the budesonide-induced genes. The most enriched functional cluster indicates positive regulation of proliferation, locomotion, movement, and migration. Moreover, comparison with the budesonide-induced expression profile in primary human airway epithelial cells shows considerable cell type specificity. In conclusion, increased expression of multiple genes, including the transcriptional repressor, ZBTB16, that reduce inflammatory signaling and gene expression, occurs in the airways and blood and may contribute to the therapeutic efficacy of ICS. This provides a previously lacking insight into the in vivo effects of ICS and should promote strategies to improve glucocorticoid efficacy in inflammatory diseases.