Application of a zero-inflated Poisson regression model of spatial distribution of viral families in bat communities across Asia

Bats are recognized as significant reservoirs for diverse viruses, many of which have zoonotic potential. Understanding these viruses' distribution and family-level associations across different bat species is critical for predicting and mitigating potential outbreaks. The Database of Bat-associated Viruses - DBatVir database is the database of zoonotic and vector-borne viruses. A zero-inflated Poisson (ZIP) regression which is a statistical model used when count data contains an excess of zero values compared to what a standard Poisson distribution would predict. A ZIP model combines two processes to generate the observed count data: The zero-generating process models the probability of observing a zero count, often using a logistic regression model. This process accounts for the excess zeros that don’t fit the Poisson data. The count process models the non-zero process using a standard Poisson distribution. Sparse data often has a high proportion of zero values. ZIP models explicitly account for this excess, making them more appropriate than standard Poisson regression. Understanding the distribution of these viruses at the family level across different bat species is essential for anticipating and preventing future outbreaks. This model effectively addressed overdispersion and excess zeros by distinguishing between true absence and low-frequency occurrences of viruses. By implementing relevant covariates from Asian bat communities, ZIP modeling allows for a more accurate and biologically meaningful interpretation of spatial and ecological patterns in viral distribution and potential zoonotic risk.