2. Thanks:
• Jukka Jernvall, Juha Laakkonen
• Barbara Tschirren
• Alan Medlar, Ari Löytynoja
• Anna Norberg, Otso Ovaskainen
• Herman Rafalinirina, Patricia Wright
Research
Foundation of
University of
Helsinki
3. Metabarcoding brings big data
• High-throughput sequencing
allows to identify species
communities in remarkable
speed
• Within-gut communities can be
identified with certain accuracy
Aivelo et al. 2018
4. Community ecology on steroids
• Large data sets can have
hundreds of samples,
thousands of ”species” and
large numbers of known
variables
• Requires statistical approach
which are also computationally
light
5. Research questions
• Which environmental or host-related variables affect within-host
community composition?
• Can we suggest potential interactions between symbionts based on
occurrence modelling?
6. Joint species modelling opportunities
Hierarchical
Modelling of Species
Communities (HMSC)
Parasite and
pathogen
phylogeny life cycle, endosymbiont,
human pathogen
geographical coordinates,
genetic similarity
parasites and microbiota as
presence/absence
environmental and host
related variables
Ovaskainen et al, 2017a; Ovaskainen et al., 2017b; Tikhonov et al., 2017
13. Possible interactions within communities
B.valaisiana
R
ickettsia
sp.
B. garinii
Rickettsia helvetica
R
. m
onacensis
Anaplasma
Lariskella
Spiroplasm
a
Rickettsiella
B. miyamotoi
C
a. N
eoehrlichia
Borreliaafzelii
Aivelo & Norberg, 2018
14. • Powerful stastistical approaches can create
testable hypothesis for species occurrence,
including species interactions
• Parasites and pathogen occurrence
determined by a range of variables
• Closely related species might have different
interactions
• Endosymbionts may play role in pathogen
invasion success
In nutshell
15. References
• Aivelo et al. 2015 https://doi.org/10.1017/S0031182015000438
• Aivelo et al. 2016 http://doi.org/10.1128/AEM.00559-16
• Aivelo et al. 2018 https://doi.org/10.1007/s10764-017-0010-x
• Aivelo & Medlar, 2018 https://doi.org/10.1017/S0031182017000610
• Aivelo & Norberg, 2016 https://doi.org/10.1101/076059
• Aivelo & Norberg, 2018 https://doi.org/10.1111/1365-2656.12708
• Medlar et al. 2014 https://doi.org/10.1186/s12862-014-0235-7
• Rafalinirina et al. 2015 https://doi.org/10.4314/mcd.v10i2.4
• Ovaskainen et al., 2017a: https://doi.org/10.1111/ele.12757
• Ovaskainen et al., 2017b: https://doi.org/10.1098/rspb.2017.0768
• Tikhonov et al., 2017: https://doi.org/10.1111/2041-210X.12723
•DOIhttps://doi.org/10.1007/s10764-017-0010-x
•DOIhttps://doi.org/10.1007/s10764-017-0010-x
Notes de l'éditeur
We did a variable selection, that is, fitted a model on which regression parameters where further modelled with each variable being 0 or 1. As a result, this builds a number of models with different sets of models. Here are pathogen OTUs and how different variables affect their prevalence, or specifically, the possibility of these OTUs being in a tick. Effects shown here are a combination of most common variables used in models
The strange part here is that B. valaisiana and B. garinii seem to be vole-affected.