During the course of this grant, we accomplished a massive expansion of the Plant Metabolic Network (PMN, https://plantcyc.org), an online database of plant metabolism. We published six major releases, bringing the number of single-species databases from 21 to 126. Each database contains enzymes predicted from the genome using our in-house pipeline E2P2, and reactions, pathways, and compounds implied by that enzyme set, along with varying degrees of curated information from the literature. PMN has added 161 metabolic pathways, 3,654 reactions, 2,777 compounds, and 1,065,384 proteins since the start of the grant. We have also added 8,705 proteins to our reference protein sequence dataset (RPSD) used to make enzyme function predictions.

PMN has been an invaluable tool to many plant biologists, with more than 1,100 unique visitors per month, 1,200 users registered for full-database downloads, and 429 literature citations. Common uses include transforming omics data, interrogating hypotheses about the evolution of plant metabolism, and annotating new genomes using the PMN BLAST feature. Our own group has published 28 papers on plant metabolism with support from this grant, including two major PMN update papers. Twenty-nine people were trained on this grant, including 6 postdocs, 6 postbac research assistants, 3 biocurators, and 14 undergraduate interns. This cohort included 15 women (52%), 15 people of color (52%), and 2 URMs (7%). All are still in STEM fields and many have moved on to the next stage of their careers, including 4 in PhD programs, 2 in MS programs, 4 in industry, 1 government, and 1 academic lab positions.  

We made significant improvements to our database-generation pipeline. The Ensemble Enzyme Prediction Pipeline (E2P2) and the associated RPSD that it uses to make predictions have been kept up to date with new information from the literature. Numerous classifiers have been tested for potential addition to E2P2, and two (the neural network-based DeepEC and the structure-based AlphaFold+TMAlign) have been selected for integration into E2P2 for future releases. We also developed a natural language processing (NLP) machine learning model to identify and assess papers with enzyme function information and differentiate between those whose enzyme classifications are based on experimental data from those based on computational prediction. We used this tool to filter enzyme function information from the BRENDA database for inclusion in the RPSD, to prevent computational predictions from being used to make more computational predictions. The semi-automated validation infrastructure (SAVI) is software that lets biocurators enter rules for inclusion or exclusion of specific pathways from the plant databases based on their phylogenetic placement. We added rules for more than 300 pathways to SAVI, bringing the total number of pathways to 1,352.

Several new website features have been implemented. Pathways that involve transport between cells or cellular compartments now show the membrane and compartments on the pathway display. Virtual PlantCyc is a new feature that allows users to select up to 10 PMN databases (genomes), for which the predicted enzymes will be pulled in, on the fly, and displayed on the PlantCyc pathway diagrams. Virtual PlantCyc also displays colored stack boxes next to each reaction of the pathway, with each box representing a genome, to indicate the presence (colored) or absence (white) of enzyme annotations from a given genome.

A number of new web applications have also been developed and published. Co-Expression Viewer has been created that can be used to view co-expression data for all the genes in a given PMN pathway, drawing data from the ATTED-II plant co-expression database. It supports all nine of the ATTED-II plant species and is accessible from PMN pathway views. Another new site shows the status of genome function annotation for several important model organisms and crops (https://genomeannotation.rheelab.org). A third new site presents plant metabolic clusters for 8 plant species computed using our PlantClusterFinder software (https://metabolicclusterviewer.dpb.carnegiescience.edu).

The project has conducted substantial outreach to the general public. We designed and implemented a program to introduce middle-school students to plant biology, and worked with Thomas R. Pollicita Middle School in Daly City, CA, a school whose student body consists of 98% Black, Indigenous and People of Color communities, to implement the program. More than 800 students across 6 classrooms learned about plant biology and the scientific process through hands-on experiments dissecting and regrowing vegetables. We have also partnered with the Canopy blog (https://canopy.org/blog) to publish 29 tree stories, posts about specific tree species, their history, and their importance to both humans and the environment. There have been 75,000 blog readers over the last 12 months and the Canopy TreEnews eblast is sent to 4,559 subscribers. The tree stories blogs authored by Rhee Lab trainees have received 41,588 views since the partnership began in 2018.

Last Modified: 10/07/2022
Modified by: Seung Rhee