Our team collaboratively develops solutions to help advance the infectious disease vaccine field. We are a strong, diverse, multidisciplinary research team with expertise in areas of statistics, mathematics, biology, bioinformatics and computing. We contribute throughout the intellectual development and execution process of research projects. We strive to overcome communication barriers that can frequently arise across disciplines, and explicitly focus on skills development in this area.
Our team includes biostatisticians with expertise and years of experience engaging with the statistical needs of vaccine and infectious disease research. We develop novel methods where they are lacking and work directly with collaborators to produce research that is thoughtful, interpretable, and reproducible.
Our team includes skilled python, R, perl, java, and C++ programmers with over two decades of experience in computational molecular biology and biological sequence analysis. We work across several major research networks to harmonize data management and analysis approaches in support of machine and human learning.
Our team includes biologists with cross-disciplinary data science expertise. Our group has extensive experience with laboratory methods and data types generated by complex biological processes and state-of-the-art technologies. We engage deeply with the science and scientists of infectious disease research.
In collaboration with CoVPN statisticians, the Edlefsen Group supports bioinformatic analysis of SARS-COV-2 breakthrough infections, and in its role as the statistical core of the United World Arbovirus Research Network (UWARN) international collaboration for emerging disease response, which has pivoted to COVID-19 research and is increasingly focused on genomic analysis of novel circulating variants of SARS-COV-2, the Edlefsen Group is engaged with local in-country researchers globally in tracking and evaluating emerging variants for vaccine escape, and for other evidence of variations relevant to the basic biology and to the public health response to the COVID-19 pandemic. Over more than one decade working in the area of data science support for translational infectious disease research, the Edlefsen Group has developed collaborations with researchers around the world, with especially strong connections with scientists studying HIV-1 in South Africa. We have successfully leveraged these relationships to expand into areas of TB, malaria, dengue, zika, and COVID-19 research.
Paul Edlefsen trained with Jun Liu (Harvard) after working for three years as a computational biologist and programmer in Lee Hood’s lab at the Institute for Systems Biology. His dissertation paper developed statistical models of transposon nucleotide alignments (profile hidden Markov models) and comparative genomics approaches for evaluating them, and in his role as Director of the HIV Vaccine Trials Network’s Computational Biology Sequence Analysis Unit, Paul leads his team in the employment of genomic analysis methodologies for evaluating impacts of treatment and prevention interventions on the genomics of infecting viruses. Sieve analysis is an application of comparative genomics across the arms of a randomized prevention trial to evaluate the efficacy as it varies by genomic features of the pathogen. The Edlefsen Group develops and applies sieve analysis methodologies for HVTN and other clinical trial networks.
In the evaluation of clinical trials for antibody mediated prophylactic interventions such as the AMP studies that evaluated impacts of VRC01 infusion on HIV-1 infection risk, the statistical analysis can critically depend on estimation of a more precise time of infection, because time-varying correlates analysis requires estimation of the time-to-event. In support of analyses of HIV-1 prevention trials such as AMP, we have developed and evaluated methodology for Bayesian combination of diagnostics-based and sequence-based estimators of HIV-1 infection time through grants funded by the Gates Foundation and NIH.
We work collaboratively with Josh Herbeck (UW, Gates IDM) on a novel modeling framework that integrates HIV-1 evolutionary processes into agent-based models of HIV epidemiology. This approach will lead to more accurate predictions of vaccine impact as it interacts with the evolving pathogen and will provide a simulation framework for evaluating statistical methods for robustness to sieve effects and variations in vaccine modality (such as leaky versus all-or-none) while accounting for subject-level exposure heterogeneities. We also work on statistical models of within-host HIV-1 evolution with applications to HIV-1 infection time estimation, with the ultimate goal of integrating these into multi-scale models for evaluation of sieve analysis methodologies that can differentiate between acquisition-blocking interventions and those that alter post-infection evolution.
The Edlefsen Group collaborates with multiple researchers working to characterize HIV-1 latency and evaluate therapeutic strategies including vaccines and antibody therapy. With Afam Okoye (OHSU) we are funded by Gilead Sciences to determine whether PD-1 blockade can enhance the functional activity of SIV-specific T cells elicited by mRNA/SIV vaccination in SIV-infected RM on cART. With Louis Picker (OHSU) we are funded by multiple sources including the Gates Foundation and the NIH to evaluate RhCMV-SIV vaccines, including for therapeutic application. With Katie Bar (U Penn) and Rebecca Lynch (GWU) we are analyzing genomic sequences pre- and post- ATI with VRC01 infusion to inform the statistical analysis plans for AMP and for ATI studies. The Edlefsen Group works with Lisa Frenkel (Seattle Children’s [SCRI]) on multiple projects evaluating HIV-1 proviral integration data from the Seattle primary infection cohort as well as from participants in the Merlin / Sabes study in Peru (with Ann Duerr, Fred Hutch), and with Jim Mullins (UW) in evaluating residual viremia towards better characterization of the latent HIV-1 reservoir in chronic infection. With Dan Barouch (Harvard) and others, the Edlefsen Group provides statistical support for a Gates-funded effort to develop models of HIV reservoirs and strategies for an HIV cure intervention. We apply the statistical and bioinformatic methods that we have developed for reservoir analysis to experimental models and human primary infection cohorts.
The Edlefsen Group designs and evaluates clinical trials for preventative vaccine evaluation and beyond. Paul Edlefsen is the lead statistician for several early-phase HVTN trials and the Edlefsen Group contributes to the design and analysis of HVTN phase IIb and phase III efficacy trials, specifically developing methods and analysis plans for sieve and correlates analyses, including HIV-1 infection timing estimation and comparative genomic sequence analysis. The Edlefsen Group contributes to the statistical design of pre-clinical trials in various model organisms, principally in non-human primates and mice. Paul Edlefsen has provided statistical support for the lab of Louis Picker (OHSU) for nearly a decade, throughout the development of the RhCMV vector vaccine for HIV, TB, and malaria. The Edlefsen Group is also the statistical core of the Cascade IMPAc-TB network and we work with Kevin Urdahl (SCRI) and collaborating labs around the world on developing models of TB vaccine responses, infection, progression, and reversion.
The Edlefsen Group is engaged in several projects involving routine and novel immunological biomarker analysis for identifying and evaluating correlates of infectious disease risk (acquisition as well as progression and reversion) across a range of disease areas, including HIV, TB, COVID-19, dengue, malaria, herpes simplex, and zika. Common data analysis workflows include ICS (stimulated T cell responses measured with flow cytometry), multiplexed ELISA, transcriptomics (bulk historically, and now increasingly single-cell), clinical cell count data (CBC), and genomic sequence measurements. We frequently work with multiple-tissue analyses across multiple labs and technologies, and we dedicate resources to constantly learning and improving the depth of our engagement with the labs employing these technologies. We are dedicated to working with data scientists across research networks to identify areas of common solutions across diseases and other foci, and seek to identify and harness synergies and cross-pollination opportunities across data management and statistical analysis efforts. Identifying and leveraging these synergies is a core strength and focus of our group, which by its breadth of collaborations and situated location within the nexus of infectious disease research is highly impactful.
Edlefsen Group members provide expertise on a wide range of projects. Together these have far-reaching impacts on infectious disease research. Our contributions to data analysis typically include custom software in R, and sometimes include custom python, perl, and other code. Our process is always open and reproducible. We maintain data and code for reproducibility in version-controlled in-house repositories for project-specific custom code, and in public repositories for software with broader applicability and to support transparent data science. We take pride in our software engineering and programming expertise. We develop tools that are useful to us and our collaborators. Our tools are research-grade and although we do not provide ongoing software support, we will work with you to help adapt our tools to specific use cases.
Paul Edlefsen trained with Art Dempster (Harvard) and his dissertation paper contributed to the recent renewed interest among mainstream statisticians in Dempster-Shafer analysis. Together with Bayesian, Fiducial and Frequentist (BFFs) practitioners from around the world, the Edlefsen Group develops statistical methodology in the foundations of statistical inference. In a landmark forthcoming paper nearly 60 years after the initial development of the Dempster-Shafer sampling procedure for categorical data, we have developed an efficient working implementation of a novel Gibbs sampling algorithm to obtain draws from the posterior distribution of bounding polytopes for use in inference. P.E. Jacob, R. Gong, P. T. Edlefsen, and A.P. Dempster. A Gibbs sampler for a class of random convex polytopes. 2021 (accepted). Journal of the American Statistical Association.