Understanding models of host-parasite coevolution

Lydia Buckingham and Ben Ashby

Our new review art­icle, ‘Coe­volu­tion­ary the­ory of hosts and para­sites’, was pub­lished in the Feb­ru­ary edi­tion of JEB. In it, we give an over­view of the the­or­et­ic­al lit­er­at­ure on host-para­site coe­volu­tion and provide key insights about mod­el assump­tions, as well as con­sid­er­ing recent devel­op­ments and future directions.

Since Haldane first sug­ges­ted that infec­tious dis­eases can drive nat­ur­al selec­tion for host res­ist­ance in 1949, a large field has developed to mod­el host-para­site coe­volu­tion the­or­et­ic­ally. These mod­els make a wide range of assump­tions about genet­ics and the envir­on­ment, and are ana­lysed using a vari­ety of tech­niques (e.g. algeb­ra­ic­ally, numer­ic­ally or through sim­u­la­tions). Com­par­is­on of such dif­fer­ent mod­els can be dif­fi­cult, so in our review we high­light the key dif­fer­ences between mod­els, explain their con­nec­tions to bio­lo­gic­al sys­tems and describe how their assump­tions affect mod­el out­comes and predictions. 

Para­sites have evolved diverse infec­tion strategies to exploit their hosts, and hosts have evolved a broad range of defence mech­an­isms to cope with para­sites. The genet­ics, eco­logy and envir­on­ment of hosts and para­sites can also vary greatly and this diversity can only be described by a wide range of dif­fer­ent mod­el­ling assumptions. 

Fig­ure 1 — A coe­volu­tion­ary “arms race”*

But which assump­tions are import­ant? Some mod­els assume that pop­u­la­tion sizes are con­stant, where­as oth­ers allow them to vary; they may assume that hosts and para­sites have over­lap­ping or non-over­lap­ping gen­er­a­tions; the genet­ic inter­ac­tions between hosts and para­sites may be mod­elled in many dif­fer­ent ways, but most can be grouped accord­ing to how they deal with the “spe­cificity” of which para­sites can infect which hosts; ran­dom effects may be included or mod­els may be determ­in­ist­ic; some­times bene­fi­cial traits are assumed to carry costs, which may either be based on empir­ic­al work or chosen arbit­rar­ily; and pop­u­la­tions may be spa­tially-struc­tured or well-mixed. 

We ana­lysed mod­els from 219 pub­lished stud­ies of host-para­site coe­volu­tion to under­stand the major dif­fer­ences between mod­els and how these affect their out­comes and pre­dic­tions. Most papers focus on when traits such as res­ist­ance and infectiv­ity will evolve to be high or low (poten­tially through a coe­volu­tion­ary “arms race” – see Fig­ure 1), when pop­u­la­tions diver­si­fy and wheth­er traits or allele fre­quen­cies will fluc­tu­ate over time (see Fig­ure 2). 

Fig­ure 2 — Fluc­tu­at­ing selec­tion between host res­ist­ance and para­site infectivity*

Our ana­lys­is of the lit­er­at­ure shows that it is import­ant to con­sider cer­tain mod­el­ling assump­tions care­fully as they can have sig­ni­fic­ant effects on mod­el out­comes. For example, fluc­tu­at­ing selec­tion tends to be sup­pressed by pop­u­la­tion dynam­ics, but is pro­moted when para­sites are highly spe­cial­ised, or when gen­er­a­tions are non-over­lap­ping. Some mod­el­ling assump­tions can also have con­trast­ing effects depend­ing on the cir­cum­stances, with stochasti­city poten­tially indu­cing fluc­tu­at­ing selec­tion (by amp­li­fy­ing damped oscil­la­tions) or dis­rupt­ing it (by caus­ing cer­tain alleles to go extinct). 

We also saw that there has been a trend towards increas­ing mod­el com­plex­ity in recent years, espe­cially in terms of more com­plex infec­tion genet­ics (e.g. multi-step pro­cesses) and inter­ac­tions with oth­er spe­cies (e.g. host-para­site-pred­at­or or host-micro­bi­o­me mod­els). Mod­els will likely con­tin­ue to become more com­plex to cap­ture more intric­ate com­munity dynam­ics or to take advant­age of new gen­om­ic data. How­ever, com­plex­ity should not be pur­sued simply for complexity’s sake, and should be groun­ded in data from real bio­lo­gic­al sys­tems where pos­sible. Cru­cially, our review high­lights the need to con­sider the effects of core mod­el­ling assump­tions so that more com­plex scen­ari­os can be prop­erly understood.

Ori­gin­al Article

Our Research Group

*© Ben Ashby. Graph­ics by Ellie Harrison.