Mapping Trypanasoma cruzi Infection Prevalence

class: center, middle, inverse, title-slide

# Mapping Trypanasoma cruzi Infection Prevalence
### Andreas Bender (@adibender)
### Nick Golding, Andre Python, Catherine Moyes
### 2019/03/18

---

# *Trypanosoma cruzi* and Chagas
American trypanosomiasis, or **Chagas disease is a leading cause of heart disease in Latin America**
- It is **one of the 10 neglected diseases addressed by the London Declaration**
which calls for **control and elimination of these devastating diseases by 2020**, based on the World Health Organization (WHO) Road map for overcoming the global impact of neglected tropical diseases.

- Caused by infection with the *Trypanosoma cruzi* parasite

- Transmitted to humans by triatomine vectors (100 species from 18 different genera)

.boxed_grey[
The primary aim of this project is to **map the prevalence of** ***T. cruzi***
**infections in vector species**, in order to investigate spatial heterogeneity
in vectorial transmission, to **target vector control**, and to
**quantify the population at risk**.
]

---
layout: true
class: split-three
.column.bg-main1[.content.vmiddle[
 
#### 1. Species distribution
<img src="figures/p_presence_raw.png">

- `$\sim$` 50 different species
- presence only data
- preferential sampling
]]
.column.bg-main2[.content[
 &nbsp;

#### 2. Infection prevalence in vectors
<img class="center", src="figures/p-spat-vect-inf-prev-1.png">
- point- and polygon-level data
- spatial sparseness

]]
.column.bg-main3[.content.vmiddle[
 
#### 3. Infection prevalence in humans
.img.fill[![](figures/p_hiprev_raw.png)]
- additional data set of acute cases
]]

# Data base

---
class: show-100

---
class: show-110
count: false

---
class: show-111
count: false

---
layout: true
class: split-three
# Road map

$$
`\begin{equation}
\eta(s) = \sum_{p=1}^{P} f_{p}(x_p(s)) + GP(s);\ p(s) = (1-\exp(\eta(s)))^{-1}
\end{equation}`
$$

.row.bg-main1[.content[
 &nbsp; &nbsp; &nbsp; &nbsp;
#### 1. Species distribution
- `$y(s) \in \{0,1\} \sim Bernoulli(p(s))$`
- **Target-group background** approach for sampling bias correction
- Stacked Generalization for combination of different methods (GAM, XGB, RF, ...)

]]
.row.bg-main2[.content[
 &nbsp;

#### 2. Infection prevalence in vectors
- `$N_{positive}(s) \sim Binomial(N_{tested}(s), p(s))$`
- SDMs for **data augmentation** (combining point-level and
polygon-level observations)
- Model evaluation based on **spatial block-wise cross-validation** to improve
transferability
]]
.row.bg-main3[.content[

#### 3. Infection prevalence in humans
- `$N_{positive}(s) \sim Binomial(N_{tested}(s), p(s))$`
- Additional **uncertainty w.r.t. location of infection** 
( `$\rightarrow$` children under-10; integration of "acute cases" data base)
]]

---
class: show-100

---
class: show-110
count: false

---
class: show-111
count: false

---
layout: true
class: split-three
.column.bg-main1[.content.vmiddle[
 
#### 1. Species distribution
.center[
<img class = "center" src="figures/p-pans-meg-gam.png" height="350">
]

]]

.column.bg-main2[.content[
 &nbsp; 
#### 2. Infection prevalence in vectors
<img class="center", src="figures/iprev-vectors.png">

]]
.column.bg-main3[.content.vmiddle[
 
#### 3. Infection prevalence in humans
<img src="figures/page_blank.jpeg">
]]

# Current progress

---
class: show-100

---
class: show-110
count: false

---
class: show-111
count: false

---
layout:false
class:
# Open Questions and Discussion

- How much does outdoor biting contribute to transmission?

- How much does contamination of food/drink by vectors contribute to transmission?

- How long after the elimination of vectors from homes (through insecticide spraying) will re-infestation occur?

- Does insecticide resistance have the potential to derail the elimination of vectors from homes?

- Does the effect of improved housing change if all houses are improved, compared to the situation where there is a mosaic of poor-quality and higher-quality housing?

---
layout: false
# References
.font80[
<a name=bib-bhatt_improved_2017></a>[Bhatt, S, E. Cameron, S. R.
Flaxman, et al.](#cite-bhatt_improved_2017) (2017). "Improved
prediction accuracy for disease risk mapping using Gaussian
process stacked generalization". En. In: _Journal of The Royal
Society Interface_ 14.134, p. 20170520. DOI:
[10.1098/rsif.2017.0520](https://doi.org/10.1098/rsif.2017.0520).

<a name=bib-browne_contemporary_2017></a>[Browne, A. J, C. A.
Guerra, R. V. Alves, et al.](#cite-browne_contemporary_2017)
(2017). "The contemporary distribution of Trypanosoma cruzi
infection in humans, alternative hosts and vectors". En. In:
_Scientific Data_ 4, p. 170050. DOI:
[10.1038/sdata.2017.50](https://doi.org/10.1038/sdata.2017.50).

<a name=bib-ceccarelli_datatri_2018></a>[Ceccarelli, S, A.
Balsalobre, P. Medone, et al.](#cite-ceccarelli_datatri_2018)
(2018). "DataTri, a database of American triatomine species
occurrence". En. In: _Scientific Data_ 5, p. 180071. DOI:
[10.1038/sdata.2018.71](https://doi.org/10.1038/sdata.2018.71).

<a name=bib-miller_recent_2019></a>[Miller, D. A. W, K. Pacifici,
J. S. Sanderlin, et al.](#cite-miller_recent_2019) (2019). "The
recent past and promising future for data integration methods to
estimate species’ distributions". En. In: _Methods in Ecology and
Evolution_ 10.1, pp. 22-37. DOI:
[10.1111/2041-210X.13110](https://doi.org/10.1111/2041-210X.13110).

<a name=bib-phillips_sample_2009></a>[Phillips, S. J, M. Dudík, J.
Elith, et al.](#cite-phillips_sample_2009) (2009). "Sample
selection bias and presence-only distribution models: implications
for background and pseudo-absence data". En. In: _Ecological
Applications_ 19.1, pp. 181-197. DOI:
[10.1890/07-2153.1](https://doi.org/10.1890/07-2153.1).

<a name=bib-roberts_cross-validation_2017></a>[Roberts, D. R, V.
Bahn, S. Ciuti, et al.](#cite-roberts_cross-validation_2017)
(2017). "Cross-validation strategies for data with temporal,
spatial, hierarchical, or phylogenetic structure". En. In:
_Ecography_ 40.8, pp. 913-929. DOI:
[10.1111/ecog.02881](https://doi.org/10.1111/ecog.02881).

<a name=bib-valavi_blockcv_2018></a>[Valavi, R, J. Elith, J. J.
Lahoz-Monfort, et al.](#cite-valavi_blockcv_2018) (2018).
"blockCV: An r package for generating spatially or environmentally
separated folds for k-fold cross-validation of species
distribution models". In: _Methods in Ecology and Evolution_ 0.0.
DOI:
[10.1111/2041-210X.13107](https://doi.org/10.1111/2041-210X.13107).
]

---
class: inverse center middle
# Appendix
<html>
 <div style='float:left'></div>
 <hr color='#EB811B' size=1px width=720px>
</html>

---
# Target-group background approach
<img src="figures/block-generation-sdm-1.png">

1. Convex hull around species presence locations

2. Extended convex hull

3. Define absences as reported presence locations of all other species

4. Set up spatial block-wise cross-validation (each fold 1-5 `$\sim$` same
  proportion of presences and absences)

---
# GAM vs. XGB
.center[
<img src="figures/p-pans-meg-gam-xgb.png" height="500px">
]