Horseshoe-based Bayesian nonparametric estimation of effective population size trajectories.

Phylodynamics is an area of population genetics that uses genetic sequence data to estimate past population dynamics. Modern state-of-the-art Bayesian nonparametric methods for recovering population size trajectories of unknown form use either change-point models or Gaussian process priors. Change-point models suffer from computational issues when the number of change-points is unknown and needs to be estimated. Gaussian process-based methods lack local adaptivity and cannot accurately recover trajectories that exhibit features such as abrupt changes in trend or varying levels of smoothness. We propose a novel, locally adaptive approach to Bayesian nonparametric phylodynamic inference that has the flexibility to accommodate a large class of functional behaviors. Local adaptivity results from modeling the log-transformed effective population size a priori as a horseshoe Markov random field, a recently proposed statistical model that blends together the best properties of the change-point and Gaussian process modeling paradigms. We use simulated data to assess model performance, and find that our proposed method results in reduced bias and increased precision when compared to contemporary methods. We also use our models to reconstruct past changes in genetic diversity of human hepatitis C virus in Egypt and to estimate population size changes of ancient and modern steppe bison. These analyses show that our new method captures features of the population size trajectories that were missed by the state-of-the-art methods.

Locally Adaptive Smoothing with Markov Random Fields and Shrinkage Priors.

We present a locally adaptive nonparametric curve fitting method that operates within a fully Bayesian framework. This method uses shrinkage priors to induce sparsity in order-k differences in the latent trend function, providing a combination of local adaptation and global control. Using a scale mixture of normals representation of shrinkage priors, we make explicit connections between our method and kth order Gaussian Markov random field smoothing. We call the resulting processes shrinkage prior Markov random fields (SPMRFs). We use Hamiltonian Monte Carlo to approximate the posterior distribution of model parameters because this method provides superior performance in the presence of the high dimensionality and strong parameter correlations exhibited by our models. We compare the performance of three prior formulations using simulated data and find the horseshoe prior provides the best compromise between bias and precision. We apply SPMRF models to two benchmark data examples frequently used to test nonparametric methods. We find that this method is flexible enough to accommodate a variety of data generating models and offers the adaptive properties and computational tractability to make it a useful addition to the Bayesian nonparametric toolbox.

Conservation planning for freshwater-marine carryover effects on Chinook salmon survival.

Experiences of migratory species in one habitat may affect their survival in the next habitat, in what is known as carryover effects. These effects are especially relevant for understanding how freshwater experience affects survival in anadromous fishes. Here, we study the carryover effects of juvenile salmon passage through a hydropower system (Snake and Columbia rivers, northwestern United States). To reduce the direct effect of hydrosystem passage on juveniles, some fishes are transported through the hydrosystem in barges, while the others are allowed to migrate in-river. Although hydrosystem survival of transported fishes is greater than that of their run-of-river counterparts, their relative juvenile-to-adult survival (hereafter survival) can be less. We tested for carryover effects using generalized linear mixed effects models of survival with over 1 million tagged Chinook salmon, Oncorhynchus tshawytscha (Walbaum) (Salmonidae), migrating in 1999-2013. Carryover effects were identified with rear-type (wild vs. hatchery), passage-type (run-of-river vs. transported), and freshwater and marine covariates. Importantly, the Pacific Decadal Oscillation (PDO) index characterizing cool/warm (i.e., productive/nonproductive) ocean phases had a strong influence on the relative survival of rear- and passage-types. Specifically, transportation benefited wild Chinook salmon more in cool PDO years, while hatchery counterparts benefited more in warm PDO years. Transportation was detrimental for wild Chinook salmon migrating early in the season, but beneficial for later season migrants. Hatchery counterparts benefited from transportation throughout the season. Altogether, wild fish could benefit from transportation approximately 2 weeks earlier during cool PDO years, with still a benefit to hatchery counterparts. Furthermore, we found some support for hypotheses related to higher survival with increased river flow, high predation in the estuary and plume areas, and faster migration and development-related increased survival with temperature. Thus, pre- and within-season information on local- and broad-scale conditions across habitats can be useful for planning and implementing real-time conservation programs.

Human-mediated evolution in a threatened species? Juvenile life-history changes in Snake River salmon.

Evaluations of human impacts on Earth's ecosystems often ignore evolutionary changes in response to altered selective regimes. Freshwater habitats for Snake River fall Chinook salmon (SRFCS), a threatened species in the US, have been dramatically changed by hydropower development and other watershed modifications. Associated biological changes include a shift in juvenile life history: Historically essentially 100% of juveniles migrated to sea as subyearlings, but a substantial fraction have migrated as yearlings in recent years. In contemplating future management actions for this species should major Snake River dams ever be removed (as many have proposed), it will be important to understand whether evolution is at least partially responsible for this life-history change. We hypothesized that if this trait is genetically based, parents who migrated to sea as subyearlings should produce faster-growing offspring that would be more likely to reach a size threshold to migrate to sea in their first year. We tested this with phenotypic data for over 2,600 juvenile SRFCS that were genetically matched to parents of hatchery and natural origin. Three lines of evidence supported our hypothesis: (i) the animal model estimated substantial heritability for juvenile growth rate for three consecutive cohorts; (ii) linear modeling showed an association between juvenile life history of parents and offspring growth rate; and (iii) faster-growing juveniles migrated at greater speeds, as expected if they were more likely to be heading to sea. Surprisingly, we also found that parents reared a full year in a hatchery produced the fastest growing offspring of all-apparently an example of cross-generational plasticity associated with artificial propagation. We suggest that SRFCS is an example of a potentially large class of species that can be considered to be "anthro-evolutionary"-signifying those whose evolutionary trajectories have been profoundly shaped by altered selective regimes in human-dominated landscapes.

Simple life-history traits explain key effective population size ratios across diverse taxa.

Effective population size (Ne) controls both the rate of random genetic drift and the effectiveness of selection and migration, but it is difficult to estimate in nature. In particular, for species with overlapping generations, it is easier to estimate the effective number of breeders in one reproductive cycle (Nb) than Ne per generation. We empirically evaluated the relationship between life history and ratios of Ne, Nb and adult census size (N) using a recently developed model (agene) and published vital rates for 63 iteroparous animals and plants. Nb/Ne varied a surprising sixfold across species and, contrary to expectations, Nb was larger than Ne in over half the species. Up to two-thirds of the variance in Nb/Ne and up to half the variance in Ne/N was explained by just two life-history traits (age at maturity and adult lifespan) that have long interested both ecologists and evolutionary biologists. These results provide novel insights into, and demonstrate a close general linkage between, demographic and evolutionary processes across diverse taxa. For the first time, our results also make it possible to interpret rapidly accumulating estimates of Nb in the context of the rich body of evolutionary theory based on Ne per generation.

Modelling evolutionary processes in small populations: not as ideal as you think.

Evolutionary processes are routinely modelled using 'ideal' Wright-Fisher populations of constant size N in which each individual has an equal expectation of reproductive success. In a hypothetical ideal population, variance in reproductive success (V(k)) is binomial and effective population size (N(e)) = N. However, in any actual implementation of the Wright-Fisher model (e.g., in a computer), V(k) is a random variable and its realized value in any given replicate generation (V(k)*) only rarely equals the binomial variance. Realized effective size (N(e)*) thus also varies randomly in modelled ideal populations, and the consequences of this have not been adequately explored in the literature. Analytical and numerical results show that random variation in V(k)* and N(e)* can seriously distort analyses that evaluate precision or otherwise depend on the assumption that N(e)* is constant. We derive analytical expressions for Var(V(k)) [4(2N - 1)(N - 1)/N(3)] and Var(N(e)) [N(N - 1)/(2N - 1) approximately N/2] in modelled ideal populations and show that, for a genetic metric G = f(N(e)), Var(G) has two components: Var(Gene) (due to variance across replicate samples of genes, given a specific N(e)*) and Var(Demo) (due to variance in N(e)*). Var(G) is higher than it would be with constant N(e) = N, as implicitly assumed by many standard models. We illustrate this with empirical examples based on F (standardized variance of allele frequency) and r(2) (a measure of linkage disequilibrium). Results demonstrate that in computer models that track multilocus genotypes, methods of replication and data analysis can strongly affect consequences of variation in N(e)*. These effects are more important when sampling error is small (large numbers of individuals, loci and alleles) and with relatively small populations (frequently modelled by those interested in conservation).

Phase II trial of carboplatin and paclitaxel in cisplatin-pretreated advanced transitional cell carcinoma: a Southwest Oncology Group study.

BACKGROUND: The purpose of the study was to assess the efficacy and toxicity of carboplatin and paclitaxel administered every 3 weeks in patients with advanced urothelial carcinoma, previously treated with cisplatin-based therapy. METHODS: Eligibility included metastatic or locally advanced unresectable transitional cell carcinoma of the urothelial tract. Prior chemotherapy, except taxanes, was permitted within 12 months. Adequate hematologic, hepatic, and renal function and a performance status of 0-2 were required. Treatment consisted of paclitaxel 200 mg/m2 intravenously for 3 hours followed by carboplatin, target area under the curve = 5 repeated every 3 weeks. RESULTS: Forty-four patients were enrolled. Thirty-four (77%) patients had a performance status of 0 or 1. Twenty-five (57%) of the patients had received prior neoadjuvant or adjuvant chemotherapy, and 19 (43%) had received it for metastatic disease. In all, 181 cycles were administered (median, 3.5 cycles; range, 1-11 cycles). The predominant NCI CTC (version 2.0) Grades 3 and 4 toxicities consisted of myelosuppression in 28 patients and neuropathy in 11 patients. There were no treatment-related deaths. Of the 44 patients, 1 (2%) had a complete response, 2 (5%) had a partial response, and 4 (9%) had an unconfirmed partial response, for an overall response rate of 16% (95% confidence interval [CI] 7-30%). The median progression-free survival was 4 months (95% CI 3-5 months) and the median survival was 6 months (95% CI 5-8 months). CONCLUSIONS: Carboplatin and paclitaxel combination is well tolerated and has modest activity in platinum refractory advanced urothelial carcinoma. Effective regimens need to be developed in cisplatin-pretreated urothelial carcinoma.

A phase II study of flavopiridol in patients with advanced renal cell carcinoma: results of Southwest Oncology Group Trial 0109.

PURPOSE: Flavopiridol is a cyclin-dependent kinase inhibitor that prevents cell cycle progression and tumor growth. In initial phase I studies, encouraging responses were seen in advanced renal cell cancer (RCC). In a phase II study of flavopiridol given as a 72-h continuous infusion every 2 weeks in RCC, a response rate of 6% was seen but with considerable grade 3 or 4 asthenia, diarrhea, and thrombosis. Subsequently, an alternative 1-h bolus schedule was reported to have enhanced tolerability in a phase I trial. We therefore conducted a phase II study of this bolus regimen. METHODS: A total of 38 patients with advanced RCC were entered into this multi-institutional phase II study. Flavopiridol (50 mg/m(2) per day) was administered by bolus intravenous injection daily for three consecutive days, repeated every 3 weeks. RESULTS: Out of 34 eligible patients, one complete response and three partial responses were observed, for an overall response rate of 12% (95% CI 3-27%). Of the 34 patients, 14 (41%) had stable disease (SD). The probability of not failing treatment by 6 months was 21% (95% CI 9-35%). Median overall survival time was 9 months (95% CI 8-18 months). The most common grade 3 or 4 toxicities were diarrhea (35%) and tumor pain (12%) along with anemia, dyspnea, and fatigue (9% each). CONCLUSIONS: Flavopiridol at this dose and schedule is feasible with an acceptable toxicity profile. Flavopiridol has some modest biologic activity against advanced RCC, as evidenced by its single-agent objective response and SD rates.

Surgical factors influence bladder cancer outcomes: a cooperative group report.

PURPOSE: A randomized, cooperative group trial (Southwest Oncology Group 8710, Intergroup 0080) reported that neoadjuvant chemotherapy improved the survival of patients with locally advanced bladder cancer who were treated with radical cystectomy. We evaluated whether surgical factors from patients enrolled onto the study predicted bladder cancer outcomes. PATIENTS AND METHODS: Surgical and tumor factors were recorded from surgical and pathologic reports from 268 patients with muscle-invasive bladder cancer who received radical cystectomy. Cystectomies were performed by 106 surgeons in 109 institutions. Half of the patients received neoadjuvant methotrexate, vinblastine, doxorubicin, and cisplatin (MVAC) chemotherapy. Variables were tested in univariate and multivariate analyses for associations with postcystectomy survival (PCS) and local recurrence (LR) in all patients receiving cystectomy. RESULTS: Five-year PCS and LR rates were 54% and 15%, respectively. A multivariate model adjusted for MVAC (P =.97), age (P =.03), pathologic stage (P =.0002), and node status (P =.04) showed that surgical variables associated with longer PCS were negative margins (v positive; hazard ratio [HR], 0.37; P =.0007), and > or = 10 nodes removed (v < 10; HR, 0.51; P =.0001). These associations did not differ by treatment arms (P >.21 for all tests of interactions between treatment and surgical variables). Predictors of LR in a multivariate model adjusted for MVAC (P =.16), pathologic stage (P =.02), and node status (P =.37) were positive margins (v negative; odds ratio [OR], 11.2; P =.0001) and fewer than 10 nodes removed (v > or = 10; OR, 5.1; P =.002). CONCLUSION: Surgical factors influence bladder cancer outcomes after cystectomy, after adjustment for pathologic factors and neoadjuvant chemotherapy usage.

Pathologic evaluation of radical cystectomy specimens: a cooperative group report.

BACKGROUND: The authors evaluated the pathology of radical cystectomy and pelvic lymph node specimens from patients with bladder carcinoma who were enrolled in a cooperative group trial. Their objective was to determine whether current practices conform to suggested pathology practice guidelines for reporting on radical cystectomy and pelvic lymph node specimens. METHODS: Overall, 268 patients underwent radical cystectomy with pelvic lymph node dissection over 11 years in a total of 109 diverse geographic locations and types of institutions. Institutions included 50 community hospitals, 36 academic hospitals, and 23 Veterans Administration (VA)/military hospitals, which evaluated 84, 137, and 47 patients, respectively. The quality of pathology examination was assessed based on the original pathology reports of specimens that were submitted according to published practice guidelines. RESULTS: Among all types of institutions, pathologic evaluation of radical cystectomy specimens generally was complete for tumor histology, grade, size, location, pathologic stage, lymph node status, prostate involvement, and associated mucosal abnormalities, including ureters and urethra. Perivesical fat (soft tissue) margins were not recorded in 10% of specimens, and 18% of patients did not mention either the presence or the number of lymph nodes. These deficiencies were observed primarily at community and VA hospitals. CONCLUSIONS: The overall quality of pathologic examination of radical cystectomy specimens is high. Better reporting of soft tissue margin status and pelvic lymph node counts is needed to achieve compliance with the standards published by evolving practice protocols. Standardized pathologic evaluation and reporting of radical cystectomy specimens will optimize important prognostic information and foster better communication between the pathologist, surgeon, and interested oncologists to benefit patients.

Ten-year survival in patients with metastatic prostate cancer.

The objective of this analysis is to identify baseline covariates that predict which patients will be long-term survivors with metastatic prostate cancer. We analyzed data from Southwest Oncology Group (SWOG) S8894, a clinical trial in men with newly diagnosed metastatic prostate cancer, to evaluate pretreatment characteristics associated with 10-year survival. There were 1286 eligible patients randomized to this study. Of these, 794 have been followed for > or = 10.5 years and are included in the analyses. Proportional odds models were used to predict 3 survival categories (survival for < 5 years, 5 up to 10 years, and > or = 10 years). Baseline patient and disease characteristics investigated were protocol treatment (flutamide vs. placebo), severity of disease, SWOG performance status (PS), bone pain, Gleason score, race, age, and prostate-specific antigen (PSA) level at study entry. Of the 794 evaluable patients, 77% lived < 5 years, 16% lived 5 up to 10 years, and 7% lived > or = 10 years. Factors predicting a statistical significant association with longer survival (P < 0.05) included minimal disease, better PS, no bone pain, lower Gleason score, and lower PSA level. All but PS were also significant in multivariate analyses. However, only 13% of patients (5 of 38) who lived > or = 10 years were correctly predicted in their survival category based on the model, whereas 98% (405 of 414) who died within the first 5 years were correctly predicted. Although statistically significant baseline characteristics were identified in this clinical trial, they did not accurately predict the survival interval to which a patient belonged.