[All-postdocs] Reminder: Bioseminar TODAY HYBRID Emelia Chamberlain and Elizabeth Connors

[Ana M Velez]

Woods Hole Oceanographic Institution
Biology Department Seminar
Thursday, March 3, 2022 - 12:00 pm HYBRID

Emelia Chamberlain
PhD student, Bowman Lab, Scripps Institution of Oceanography

Identifying Microbial Drivers of Biological Oxygen Production and Uptake in the Central Arctic Ocean
There is considerable uncertainty in the net trophic status of the Arctic Ocean, with previous work showing regional dependence and influences of both physical and biological processes. Observations of microbial community structure serve as a potential tool for estimating trophic status due to the intimate links between biogeochemical turnover and microbial metabolism. It is increasingly important to understand and quantify these ecological connections as rapid climatic changes cascade through the system with unpredictable effects. The 2019-2020 MOSAiC International Arctic Drift Expedition provided the opportunity to collect paired productivity and diversity data across the Arctic seasonal cycle. I'll present preliminary results from the last 7 months of the MOSAiC Drift, showing water column microbial community structure derived from 16S/18S rRNA gene sequencing and biological oxygen utilization derived from O2/Ar excursions analyzed using a shipboard Membrane Inlet Mass Spectrometer. To link the two measurements, a random forest regression is used to identify key microbial predictors of changes in net trophic status. We expect this work to elucidate diversity driven mechanisms for biogeochemical processes and test new diagnostics to improve models of ecological functioning in a changing Arctic.

Elizabeth Connors
PhD student, Bowman Lab, Scripps Institution of Oceanography

Using Machine Learning to Predict Bacterial Production from Absolute Prokaryotic Abundance in Waters Along the Western Antarctic Peninsula
The coastal Western Antarctic Peninsula (WAP) is a highly productive ecosystem where bacterial productivity (BP) has shown to vary over depths, locations, and years. Better predictions of the transfer of dissolved organic matter via the microbial food web to higher trophic levels is ecologically significant as it directly impacts nutrient cycling, primary production and large-scale ecosystem functions. We applied random forests, a widely used supervised machine learning algorithm, to predict BP from prokaryotic community structure data. In collaboration with the Palmer Long Term Ecological Research (LTER) project, 480 samples were collected for bacterial community structure (via 16S rRNA gene sequencing) and bacterial abundance and cell size (via flow cytometry) over five recent field seasons (2015-2020). Measurements for BP were calculated via leucine incorporation for 125 of these samples (26%). From the 16S rRNA gene sequences, we determined community structure via the paprica pipeline, and the relative abundances of each amplicon sequence variant were converted to an estimated absolute value by multiplying by the flow cytometric cell counts of each sample. Those samples with BP data were split into training and testing datasets, and a random forest regression with feature selection was carried out on the training data. The resulting model predicted BP on the test dataset with high fidelity (R2 = 0.81). We then constructed a new model to predict bacterial production for those samples where BP data were missing, thus expanding our understanding of how BP is distributed in space and time along the WAP, and which microbial taxa are the key drivers of BP in this ecosystem.

In person: Redfield Auditorium.
Zoom https://whoi-edu.zoom.us/j/98638413781 Meeting ID: 986 3841 3781 Passcode: JW.pd0
By dial: Find your local number: https://whoi-edu.zoom.us/u/abfiBkQ4df Passcode: 800440
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.whoi.edu/pipermail/all-postdocs/attachments/20220303/69db80c4/attachment.htm