Chin J Plant Ecol ›› 2019, Vol. 43 ›› Issue (7): 611-623.DOI: 10.17521/cjpe.2019.0065

• Research Articles • Previous Articles     Next Articles

Selection of null models in nestedness pattern detection of highly asymmetric mycorrhizal networks

LIN Li-Tao,MA Ke-Ming()   

  1. State Key Laboratory of Urban and Regional Ecology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085,China;and University of Chinese Academy of Sciences, Beijing 100049, China
  • Received:2019-03-25 Accepted:2019-06-05 Online:2019-07-20 Published:2019-12-12
  • Contact: MA Ke-Ming
  • Supported by:
    Supported by the National Natural Science Foundation of China(31470481)


Aims Null model is an important basis for nesting judgment. Highly asymmetric structures often appear in plant symbolic fungal networks. This study aims to explore the influence of matrix asymmetric changes on network nesting judgment.
Methods The study was conducted based on various null model construction methods.
Important findings Constraints vary with changing null models, with reducing null space when additional qualifications were added during null model establishment. Highly constrained nulls are prone to causing type II errors. Highly asymmetric networks increase matrix temperature (NT) deviation based on random (Equiprobable- equiprobable, r00) null model while reducing overlap and decreasing fill (NODF) deviation. Values of z-score show that highly asymmetric networks contribute to the significant determination level of NT and NODF. The impacts on the judgment of nestedness of asymmetric networks differ between row and column fixed null models. The effects of network asymmetry change on nesting detection based on column constrained (c0) nulls are similar to that of random null model, but with smaller nesting deviation and standard deviations. No significant differences in both NT and NT deviations were observed among different asymmetry networks based on the row fixed (r0) nulls, with a lower NODF deviation in highly asymmetric network based on c0 nulls. To more accurately determine whether the asymmetric networks would have nested structures, we recommend using a combination of random and constrained null models. Our results also demonstrate that the r0 null model performs better than either the r00 null model or the c0 null model when comparing nesting level of different asymmetric networks.

Key words: interaction network, nestedness, null model, web asymmetry