ncov-variant-rt

About

Analysis of state-by-variant Rt estimates using CDC and GISAID data

Eslam Abousamra, Marlin Figgins, John Huddleston, Trevor Bedford

This project involves a demonstration to build out a framework to perform measurements of the effective reproductive number (Rt) using bayesian inference methods of estimation, mostly involving Rstudio epidemiological tool packages which include EpiEstim (Cori A. et al., 2013), EpiNow (Abbott S. et al., 2020). In addition to questioning the effect of interventions such as vaccination and address prediction limitations and biases.

Tidying Data

Installation

Install R and then install the following R packages.

install.packages("tidyverse")
install.packages("openintro")

Install Nextstrain and the ncov workflow. These steps should provide you with a Conda environment named nextstrain and a git repository named ncov in the parent directory of this project (e.g., ../ncov/).

Initial datasets

GISAID metadata

The SARS-CoV-2 genome data were obtained from the GISAID Initiative, please refer to https://www.gisaid.org for more information. To get access to the metadata dataset, register for a free GISAID account.

Download and sanitize the metadata as described in the SARS-CoV-2 data preparation guide.

python3 ../ncov/scripts/sanitize_metadata.py \
    --metadata metadata_tsv_2021_10_28.tar.xz \
    --database-id-columns "Accession ID" \
    --parse-location-field Location \
    --rename-fields 'Virus name=strain' 'Accession ID=gisaid_epi_isl' 'Collection date=date' 'Pango lineage=pango_lineage' \
    --strip-prefixes "hCoV-19/" \
    --output metadata_gisaid.tsv.gz

After sanitizing the GISAID metadata into a file named metadata_gisaid.tsv.gz, select all records from the USA.

augur filter \
  --metadata metadata_gisaid.tsv.gz \
  --min-date 2021-01-01 \
  --query "country == 'USA'" \
  --output-metadata metadata_gisaid_usa_since_2021-01-01.tsv.gz

CDC case counts

The SARS-CoV-2 daily case counts were obtained from the CDC's Data service. Navigate to the United States COVID-19 Cases and Deaths by State over Time. Select "Export" from the top-right menu and then select "CSV" as the format to download. Save this file as "cdc-data.csv" in the top-level project directory.

Alternately, download the data directly from the command line.

curl -o cdc-data.csv "https://data.cdc.gov/api/views/9mfq-cb36/rows.csv?accessType=DOWNLOAD"

Exploratory Analysis

Here we fit a multinomial logistic regression on SARS-CoV-2 variant frequency growth (estimated from the GISAID metadata) and perfoming initial exploratory analysis and visualizing the competition among the lineages within divisions in the US.

Merging and Tidying the dataset

1. Navigate in R the code found in the Rmd file below to merge the desired dataset (CDC & GISAID)

Rt_dataset_tidy.Rmd

2. Output tidy dataset as

write.csv(initial, "counts_and_frequencies_per_date_state_and_lineage.csv")

Initial Analysis: Smoothing frequency (Multinomial Logistic Regression)

1. Navigate in R the code found in the Rmd file below to input tidy dataset and perform the MLR

Rt_MLR_tidy.Rmd

Visualizing smoothed frequencies for lineages of concern (B.1.1.7, B.1.351, P.1, B.1.427, B.1.617)

list obtained from the cdc and can be altered according to the desired analysis, please refer to https://www.cdc.gov/coronavirus/2019-ncov/variants/variant-info.html for a complete SARS-CoV-2 Variant Classifications

Visualizing Cross Validation of smoothed frequencies

2. Output tidy dataset with regression parameters to pass in the estimation models as

write.csv(final_mod, "Rt_MLR_cdc&GISAID.csv")

Rt Estimation Analysis

EpiEstim

Here we quantify transmissibility throughout an epidemic from the analysis of time series of incidence as described in Cori et al. (2013) on SARS-CoV-2 data.

Citation: Anne Cori, Neil M. Ferguson, Christophe Fraser, Simon Cauchemez, A New Framework and Software to Estimate Time-Varying Reproduction Numbers During Epidemics, American Journal of Epidemiology, Volume 178, Issue 9, 1 November 2013, Pages 1505–1512.

Installation

install.packages("EpiEstim")

Model Implementation

1. Navigate in R the code found in the Rmd file below to run the analysis for the tidy data

06-28-21-Rt_EpiEstim_1.Rmd

Visualizing Rt estimated from the EpiEstim package on a number of states

EpiNow

Using this package, we can estimate Rt through a Bayesian variable approach using the probabilistic programming language Stan. According to a gaussian process, the fitted model of Rt is an multivariate normal distribution function.

Citation: Abbott SHellewell J, Sherratt K, Gostic K, Hickson J, Badr HS, DeWitt M, Thompson R, EpiForecasts , Funk S. 2020EpiNow2: estimate real-time case counts and time-varying epidemiological parameters.

Installation

install.packages("EpiNow2")

Model Implementation

1. Navigate in R the code found in the Rmd file below to run the analysis for the tidy data

07-01-21-Rt_EpiNow_2.Rmd