F. William Townes
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
F. William Townes has a total of 43 co-authors. Here are the top co-authors by number of shared publications:
- Rafael A. Irizarry
Department of Data Science, Dana-Farber Cancer Institute, Boston, MA, USA
7 shared publications - Stephanie C. Hicks
Department of Biostatistics, Johns Hopkins University
5 shared publications - Barbara Elizabeth Engelhardt
Stanford University, USA
4 shared publications - Martin J. Aryee
Department of Data Science, Dana-Farber Cancer Institute
3 shared publications - Andrew Jones
Department of Computer Science, Princeton University
2 shared publications - Gabrielle Thivierge
Department of Statistics & Data Science, Carnegie Mellon University, Pittsburgh, 15213, PA, USA
2 shared publications
Latest Publications
- Open Access
Longitudinal single-cell modeling reveals monocyte reprogramming in juvenile systemic sclerosis following autologous stem cell transplantation
Juvenile systemic sclerosis (jSSc) is a rare autoimmune disease marked by skin fibrosis and multi-organ involvement. Autologous stem cell transplantation (ASCT) is an emerging therapy for severe, treatment-refractory jSSc, but its effects on immune cell dynamics remain poorly understood. PBMCs were collected from three patients with jSSc before ASCT and at 6, 12, and 24 months post-ASCT. Patient and healthy control samples were profiled using cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq). We focused on monocytes, given their role in fibrosis-promoting inflammation. To detect longitudinal trends, pseudobulked gene expression (log scale) was regressed against time since ASCT. This approach identified widespread changes in jSSc monocytes, including decreased expression of systemic sclerosis-linked genes, such as SERPINE1. On the pathway level, NF-κB-associated inflammatory signaling was elevated in jSSc monocytes at baseline relative to healthy controls and decreased progressively post-ASCT. Genes related to mitochondrial function and oxidative phosphorylation progressively increased in expression after ASCT, suggesting a shift in metabolic state. Compositional changes in monocyte subpopulations were also identified and may have contributed to longitudinal gene expression patterns. Together, these findings characterize the dynamic immune changes in jSSc following ASCT and highlight a widely applicable longitudinal modeling framework for single-cell data.
Open AccessLongitudinal single-cell modeling reveals monocyte reprogramming in juvenile systemic sclerosis following autologous stem cell transplantation
Juvenile systemic sclerosis (jSSc) is a rare autoimmune disease marked by skin fibrosis and multi-organ involvement. Autologous stem cell transplantation (ASCT) is an emerging therapy for severe, treatment-refractory jSSc, but its effects on immune cell dynamics remain poorly understood. PBMCs were collected from three patients with jSSc before ASCT and at 6, 12, and 24 months post-ASCT. Patient and healthy control samples were profiled using cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq). We focused on monocytes, given their role in fibrosis-promoting inflammation. To detect longitudinal trends, pseudobulked gene expression (log scale) was regressed against time since ASCT. This approach identified widespread changes in jSSc monocytes, including decreased expression of systemic sclerosis-linked genes, such as SERPINE1. On the pathway level, NF-κB-associated inflammatory signaling was elevated in jSSc monocytes at baseline relative to healthy controls and decreased progressively post-ASCT. Genes related to mitochondrial function and oxidative phosphorylation progressively increased in expression after ASCT, suggesting a shift in metabolic state. Compositional changes in monocyte subpopulations were also identified and may have contributed to longitudinal gene expression patterns. Together, these findings characterize the dynamic immune changes in jSSc following ASCT and highlight a widely applicable longitudinal modeling framework for single-cell data.
- 1 January 2026
- Published by Institute of Mathematical Statistics in Electronic Communications in Probability
- Vol. 31 (none) , 1-13
Mixed Poisson families with real-valued mixing distributions
Mixed Poisson distributions provide a flexible approach to the analysis of count data with overdispersion, zero inflation, or heavy tails. Since the Poisson mean must be nonnegative, the mixing distribution is typically assumed to have nonnegative support. We show this assumption is unnecessary and real-valued mixing distributions are also possible. Informally, the mixing distribution merely needs to have a light (decaying faster than exponential) left tail and a small amount of probability mass on negative values. We provide several concrete examples, including the mixed Poisson- extreme stable family, where the mixing distribution has a power law tail.
- 1 January 2026
- Published by Institute of Mathematical Statistics in Electronic Communications in Probability
- Vol. 31 (none) , 1-13
Mixed Poisson families with real-valued mixing distributions
Mixed Poisson distributions provide a flexible approach to the analysis of count data with overdispersion, zero inflation, or heavy tails. Since the Poisson mean must be nonnegative, the mixing distribution is typically assumed to have nonnegative support. We show this assumption is unnecessary and real-valued mixing distributions are also possible. Informally, the mixing distribution merely needs to have a light (decaying faster than exponential) left tail and a small amount of probability mass on negative values. We provide several concrete examples, including the mixed Poisson- extreme stable family, where the mixing distribution has a power law tail.
- Open Access
Broadly discrete stable distributions
Stable distributions are of fundamental importance in probability theory, yet their absolute continuity makes them unsuitable for modeling count data. A discrete analog of strict stability has been previously proposed by replacing scaling with binomial thinning, but it only holds for a subset of the tail index parameters. Here, we generalize the discrete stable class to the full range of tail indices and show that it is equivalent to the mixed Poisson-stable family. This broadly discrete stable family is discretely infinitely divisible, with a compound Poisson representation involving a novel generalization of the Sibuya distribution. Under additional parameter constraints, they are also discretely self-decomposable and unimodal. The discrete stable distributions provide a new frontier in probabilistic modeling of both light and heavy tailed count data.
Open AccessBroadly discrete stable distributions
Stable distributions are of fundamental importance in probability theory, yet their absolute continuity makes them unsuitable for modeling count data. A discrete analog of strict stability has been previously proposed by replacing scaling with binomial thinning, but it only holds for a subset of the tail index parameters. Here, we generalize the discrete stable class to the full range of tail indices and show that it is equivalent to the mixed Poisson-stable family. This broadly discrete stable family is discretely infinitely divisible, with a compound Poisson representation involving a novel generalization of the Sibuya distribution. Under additional parameter constraints, they are also discretely self-decomposable and unimodal. The discrete stable distributions provide a new frontier in probabilistic modeling of both light and heavy tailed count data.
- Open Access
- 16 December 2024
- Published by American Chemical Society (ACS) in ACS ES&T Water
- Vol. 5 (1) , 341-350
Spatiotemporal Variability of the Pepper Mild Mottle Virus Biomarker in Wastewater
Since the start of the coronavirus-19 pandemic, the use of wastewater-based epidemiology (WBE) for disease surveillance has increased throughout the world. Because wastewater measurements are affected by external factors, processing WBE data typically includes a normalization step in order to adjust wastewater measurements (e.g., viral ribonucleic acid (RNA) concentrations) to account for variation due to dynamic population changes, sewer travel effects, or laboratory methods. Pepper mild mottle virus (PMMoV), a plant RNA virus abundant in human feces and wastewater, has been used as a fecal contamination indicator and has been used to normalize wastewater measurements extensively. However, there has been little work to characterize the spatiotemporal variability of PMMoV in wastewater, which may influence the effectiveness of PMMoV for adjusting or normalizing WBE measurements. Here, we investigate its variability across space and time using data collected over a two-year period from sewage treatment plants across the United States. We find that most variation in PMMoV measurements can be attributed to longitude and latitude followed by site-specific variables. Further research into cross-geographical and -temporal comparability of PMMoV-normalized pathogen concentrations would strengthen the utility of PMMoV in WBE.
- 16 December 2024
- Published by American Chemical Society (ACS) in ACS ES&T Water
- Vol. 5 (1) , 341-350
Open AccessSpatiotemporal Variability of the Pepper Mild Mottle Virus Biomarker in Wastewater
Since the start of the coronavirus-19 pandemic, the use of wastewater-based epidemiology (WBE) for disease surveillance has increased throughout the world. Because wastewater measurements are affected by external factors, processing WBE data typically includes a normalization step in order to adjust wastewater measurements (e.g., viral ribonucleic acid (RNA) concentrations) to account for variation due to dynamic population changes, sewer travel effects, or laboratory methods. Pepper mild mottle virus (PMMoV), a plant RNA virus abundant in human feces and wastewater, has been used as a fecal contamination indicator and has been used to normalize wastewater measurements extensively. However, there has been little work to characterize the spatiotemporal variability of PMMoV in wastewater, which may influence the effectiveness of PMMoV for adjusting or normalizing WBE measurements. Here, we investigate its variability across space and time using data collected over a two-year period from sewage treatment plants across the United States. We find that most variation in PMMoV measurements can be attributed to longitude and latitude followed by site-specific variables. Further research into cross-geographical and -temporal comparability of PMMoV-normalized pathogen concentrations would strengthen the utility of PMMoV in WBE.
- Open Access
Does Spatial Information Improve Influenza Forecasting?
Seasonal influenza forecasting is critical for public health and individual decision making. We investigate whether the inclusion of data about influenza activity in neighboring states can improve point predictions and distribution forecasting of influenza-like illness (ILI) in each US state using statistical regression models. Using CDC FluView ILI data from 2010-2019, we forecast weekly ILI in each US state with quantile, linear, and Poisson autoregressive models fit using different combinations of ILI data from the target state, neighboring states, and US weighted average. Scoring with root mean squared error and weighted interval score indicated that the variants including neighbors and/or the US average showed slightly higher accuracy than models fit only using lagged ILI in the target state, on average. Additionally, the improvement in performance when including neighbors was similar to the improvement when including the US average instead, suggesting the proximity of the neighboring states is not the driver of the slight increase in accuracy.
Open AccessDoes Spatial Information Improve Influenza Forecasting?
Seasonal influenza forecasting is critical for public health and individual decision making. We investigate whether the inclusion of data about influenza activity in neighboring states can improve point predictions and distribution forecasting of influenza-like illness (ILI) in each US state using statistical regression models. Using CDC FluView ILI data from 2010-2019, we forecast weekly ILI in each US state with quantile, linear, and Poisson autoregressive models fit using different combinations of ILI data from the target state, neighboring states, and US weighted average. Scoring with root mean squared error and weighted interval score indicated that the variants including neighbors and/or the US average showed slightly higher accuracy than models fit only using lagged ILI in the target state, on average. Additionally, the improvement in performance when including neighbors was similar to the improvement when including the US average instead, suggesting the proximity of the neighboring states is not the driver of the slight increase in accuracy.
- Open Access
Exponential tilting of subweibull distributions
The class of subweibull distributions has recently been shown to generalize the important properties of subexponential and subgaussian random variables. We describe alternative characterizations of subweibull distributions and detail the conditions under which their tail behavior is preserved after exponential tilting.
Open AccessExponential tilting of subweibull distributions
The class of subweibull distributions has recently been shown to generalize the important properties of subexponential and subgaussian random variables. We describe alternative characterizations of subweibull distributions and detail the conditions under which their tail behavior is preserved after exponential tilting.
- Open Access
- 17 August 2023
- Published by Springer Nature in Nature Methods
- Vol. 20 (9) , 1379-1387
Alignment of spatial genomics data using deep Gaussian processes
Spatially resolved genomic technologies have allowed us to study the physical organization of cells and tissues, and promise an understanding of local interactions between cells. However, it remains difficult to precisely align spatial observations across slices, samples, scales, individuals and technologies. Here, we propose a probabilistic model that aligns spatially-resolved samples onto a known or unknown common coordinate system (CCS) with respect to phenotypic readouts (for example, gene expression). Our method, Gaussian Process Spatial Alignment (GPSA), consists of a two-layer Gaussian process: the first layer maps observed samples’ spatial locations onto a CCS, and the second layer maps from the CCS to the observed readouts. Our approach enables complex downstream spatially aware analyses that are impossible or inaccurate with unaligned data, including an analysis of variance, creation of a dense three-dimensional (3D) atlas from sparse two-dimensional (2D) slices or association tests across data modalities.
- 17 August 2023
- Published by Springer Nature in Nature Methods
- Vol. 20 (9) , 1379-1387
Open AccessAlignment of spatial genomics data using deep Gaussian processes
Spatially resolved genomic technologies have allowed us to study the physical organization of cells and tissues, and promise an understanding of local interactions between cells. However, it remains difficult to precisely align spatial observations across slices, samples, scales, individuals and technologies. Here, we propose a probabilistic model that aligns spatially-resolved samples onto a known or unknown common coordinate system (CCS) with respect to phenotypic readouts (for example, gene expression). Our method, Gaussian Process Spatial Alignment (GPSA), consists of a two-layer Gaussian process: the first layer maps observed samples’ spatial locations onto a CCS, and the second layer maps from the CCS to the observed readouts. Our approach enables complex downstream spatially aware analyses that are impossible or inaccurate with unaligned data, including an analysis of variance, creation of a dense three-dimensional (3D) atlas from sparse two-dimensional (2D) slices or association tests across data modalities.
- Top Cited PapersOpen Access
- 31 December 2022
- Published by Springer Nature in Nature Methods
- Vol. 20 (2) , 229-238
Nonnegative spatial factorization applied to spatial genomics
Nonnegative matrix factorization (NMF) is widely used to analyze high-dimensional count data because, in contrast to real-valued alternatives such as factor analysis, it produces an interpretable parts-based representation. However, in applications such as spatial transcriptomics, NMF fails to incorporate known structure between observations. Here, we present nonnegative spatial factorization (NSF), a spatially-aware probabilistic dimension reduction model based on transformed Gaussian processes that naturally encourages sparsity and scales to tens of thousands of observations. NSF recovers ground truth factors more accurately than real-valued alternatives such as MEFISTO in simulations, and has lower out-of-sample prediction error than probabilistic NMF on three spatial transcriptomics datasets from mouse brain and liver. Since not all patterns of gene expression have spatial correlations, we also propose a hybrid extension of NSF that combines spatial and nonspatial components, enabling quantification of spatial importance for both observations and features. A TensorFlow implementation of NSF is available from https://github.com/willtownes/nsf-paper.
- 31 December 2022
- Published by Springer Nature in Nature Methods
- Vol. 20 (2) , 229-238
Top Cited PapersOpen AccessNonnegative spatial factorization applied to spatial genomics
Nonnegative matrix factorization (NMF) is widely used to analyze high-dimensional count data because, in contrast to real-valued alternatives such as factor analysis, it produces an interpretable parts-based representation. However, in applications such as spatial transcriptomics, NMF fails to incorporate known structure between observations. Here, we present nonnegative spatial factorization (NSF), a spatially-aware probabilistic dimension reduction model based on transformed Gaussian processes that naturally encourages sparsity and scales to tens of thousands of observations. NSF recovers ground truth factors more accurately than real-valued alternatives such as MEFISTO in simulations, and has lower out-of-sample prediction error than probabilistic NMF on three spatial transcriptomics datasets from mouse brain and liver. Since not all patterns of gene expression have spatial correlations, we also propose a hybrid extension of NSF that combines spatial and nonspatial components, enabling quantification of spatial importance for both observations and features. A TensorFlow implementation of NSF is available from https://github.com/willtownes/nsf-paper.
- 1 September 2022
- Published by Institute of Mathematical Statistics in The Annals of Applied Statistics
- Vol. 16 (3) , 1268-1291
Contrastive latent variable modeling with application to case-control sequencing experiments
High-throughput RNA-sequencing (RNA-seq) technologies are powerful tools for understanding cellular state. Often, it is of interest to quantify and to summarize changes in cell state that occur between experimental or biological conditions. Differential expression is typically assessed using univariate tests to measure genewise shifts in expression. However, these methods largely ignore changes in transcriptional correlation. Furthermore, there is a need to identify the low-dimensional structure of the gene expression shift to identify collections of genes that change between conditions. Here, we propose contrastive latent variable models designed for count data to create a richer portrait of differential expression in sequencing data. These models disentangle the sources of transcriptional variation in different conditions in the context of an explicit model of variation at baseline. Moreover, we develop a model-based hypothesis testing framework that can test for global and gene subset-specific changes in expression. We evaluate our model through extensive simulations and analyses with count-based gene expression data from perturbation and observational sequencing experiments. We find that our methods effectively summarize and quantify complex transcriptional changes in case-control experimental sequencing data.
- 1 September 2022
- Published by Institute of Mathematical Statistics in The Annals of Applied Statistics
- Vol. 16 (3) , 1268-1291
Contrastive latent variable modeling with application to case-control sequencing experiments
High-throughput RNA-sequencing (RNA-seq) technologies are powerful tools for understanding cellular state. Often, it is of interest to quantify and to summarize changes in cell state that occur between experimental or biological conditions. Differential expression is typically assessed using univariate tests to measure genewise shifts in expression. However, these methods largely ignore changes in transcriptional correlation. Furthermore, there is a need to identify the low-dimensional structure of the gene expression shift to identify collections of genes that change between conditions. Here, we propose contrastive latent variable models designed for count data to create a richer portrait of differential expression in sequencing data. These models disentangle the sources of transcriptional variation in different conditions in the context of an explicit model of variation at baseline. Moreover, we develop a model-based hypothesis testing framework that can test for global and gene subset-specific changes in expression. We evaluate our model through extensive simulations and analyses with count-based gene expression data from perturbation and observational sequencing experiments. We find that our methods effectively summarize and quantify complex transcriptional changes in case-control experimental sequencing data.