Ecological Network Inference From Long-Term Presence-Absence Data

Ecological communities are characterized by complex networks of trophic and nontrophic interactions, which shape the dy-namics of the community. Machine learning and correlational methods are increasingly popular for inferring networks from co-occurrence and time series data, particularly in microbial systems. In this study, we test the suitability of these methods for inferring ecological interactions by constructing networks using Dynamic Bayesian Networks, Lasso regression, and Pear-son’s correlation coefficient, then comparing the model networks to empirical trophic and nontrophic webs in two ecological systems. We find that although each model significantly replicates the structure of at least one empirical network, no model significantly predicts network structure in both systems, and no model is clearly superior to the others. We also find that networks inferred for the Tatoosh intertidal match the nontrophic network much more closely than the trophic one, possibly due to the challenges of identifying trophic interactions from presence-absence data. Our findings suggest that although these methods hold some promise for ecological network inference, presence-absence data does not provide enough signal for models to consistently identify interactions, and networks inferred from these data should be interpreted with caution.


True Positive
False Positive False Negative Figure S1. Structural comparison between empirical and model adjacency matrices as in Fig. 3a, but containing all species used in the DBN-2 model.

True Positive
False Positive False Negative Figure S2. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the France piscivory network and the DBN-3 model.

True Positive
False Positive False Negative Figure S3. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the France piscivory network and the DBN-4 model.

True Positive
False Positive False Negative Figure S4. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the France piscivory network and the DBN-5 model.

True Positive
False Positive False Negative Figure S5. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the France piscivory network and the Lasso-1st model.

True Positive
False Positive False Negative Figure S6. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the France piscivory network and the Lasso-2nd model.

True Positive
False Positive False Negative Figure S7. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the France piscivory network and the Pearson model.

True Positive
False Positive False Negative Figure S8. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the Tatoosh trophic network and the DBN-2 model. Identical to Fig. 3a, except that it contains all species in the DBN and Pearson models, rather than excluding those which were excluded from the Lasso models.

True Positive
False Positive False Negative Figure S9. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the Tatoosh trophic network and the DBN-3 model. Contains all species used in the DBN and Pearson models.

True Positive
False Positive False Negative Figure S10. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the Tatoosh trophic network and the DBN-4 model. Contains all species used in the DBN and Pearson models.

True Positive
False Positive False Negative Figure S11. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the Tatoosh trophic network and the DBN-5 model. Contains all species used in the DBN and Pearson models.

True Positive
False Positive False Negative Figure S12. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the Tatoosh trophic network and the Lasso-2nd model.

True Positive
False Positive False Negative Figure S13. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the Tatoosh trophic network and the Lasso-2nd model. Identical to Fig. 3e, except that it contains all species in the DBN and Pearson models, rather than excluding those which were excluded from the Lasso models.

True Positive
False Positive False Negative Figure S14. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the Tatoosh nontrophic network and the DBN-2 model. Identical to Fig. 3b, except that it contains all species in the DBN and Pearson models, rather than excluding those which were excluded from the Lasso models.

19/29
True False Negative Figure S18. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the Tatoosh nontrophic network and the Lasso-2nd model.

True Positive
False Positive False Negative Figure S19. Structural comparison between empirical and model adjacency matrices as in Fig. 3, but for the Tatoosh nontrophic network and the Pearson model. Identical to Fig. 3f, except that it contains all species in the DBN and Pearson models, rather than excluding those which were excluded from the Lasso models.  Table S1. Expanded table of precision and recall with p-values. Columns represent the empirical network used, the model used to construct the inferred network, and the FDR-corrected p-values, rounded to two decimal places.