Using machine learning to associate bacterial taxa with functional groups through flow cytometry, 16S rRNA gene sequencing, and productivity data

Abstract: High- (HNA) and low-nucleic acid (LNA) bacteria are two separated flow cytometry (FCM) groups that are ubiquitous across aquatic systems. HNA cell density often correlates strongly with heterotrophic production. However, the taxonomic composition of bacterial taxa within HNA and LNA groups remains mostly unresolved. Here, we associated freshwater bacterial taxa with HNA and LNA groups by integrating FCM and 16S rRNA gene sequencing using a machine learning-based variable selection approach. There was a strong association between bacterial heterotrophic production and HNA cell abundances (R2 = 0.65), but not with more abundant LNA cells, suggesting that the smaller pool of HNA bacteria may play a disproportionately large role in the freshwater carbon flux. Variables selected by the models were able to predict HNA and LNA cell abundances at all taxonomic levels, with highest accuracy at the OTU level. There was high system specificity as the selected OTUs were mostly unique to each lake ecosystem and some OTUs were selected for both groups or were rare. Our approach allows for the association of OTUs with FCM functional groups and thus the identification of putative indicators of heterotrophic activity in aquatic systems, an approach that can be generalized to other ecosystems and functioning of interest.

Authors: Peter Rubbens, Marian L. Schmidt, Ruben Props, Bopaiah A. Biddanda, Nico Boon, Willem Waegeman, Vincent J. Denef

Peter Rubbens and Marian L. Schmidt contributed equally to this work.

Variable selection

Examples of how variable selection was performed using the Randomized Lasso and Boruta algorithm are given in two Jupyter notebooks: fs_Muskegon_HNA_5seq10.ipynb and fs_Muskegon_LNA_5seq10.ipynb. Scripts to perform recursive variable elimination for these cases can be found in RFE_HNA_MUS_5seq10_RFE_Lasso.py and RFE_LNA_MUS_5seq10_RFE_Lasso.py. Figures 2 and 3 can be generated with the plot_Fig2.py and plot_Fig3.py scripts. Functions that are used can be found in the file analysis_functions.py.

Results

Variable selection results for all all lake systems are stored in the directory FS_final. Results based on a recursive variable elimination can be found in the directory RFE.

Dependencies

• Numpy: Scientific Computing with Python.
• Pandas: Python Data Structure and Analysis Library.
• Scikit-Learn: Python Machine Learning Library.
• Boruta_py: Pythom implementation of the Boruta algorithm.

Phylogenetic Signal

Please click here for the code for Phylogenetic Signal.

Phylogenetic Tree Construction

The phylogenetic tree in Figure 5 was calculated using fasttree and was visualized in iTOL. Data files imported for tree construction can be found in data/Figure5.

Figures 1, 3, and 4

Please click here for the code for Figures 1, 3, and 4.

Supplemental Information

The supplemental tables and figures can be found in supplemental_info/.

For a supplemental analysis testing which groups of OTUs correlate with bacterial production using weighted gene correlation network analysis (WGCNA), please see this link.