Improving the Efficacy of the Tracer Algorithm as a Gap-Filling and Pathway Prediction Tool for Metabolic Models
Jared Coffelt
Author
07/29/2020
Added
45
Plays
Description
The Tracer Gap-Filling Algorithm is a tool designed to fill gaps in pathways in metabolic models. Still in developmental phases, research focused on improving the performance of the model and validation of its results.
Searchable Transcript
Toggle between list and paragraph view.
- [00:00:02.532]Hello, my name is Jared Coffelt
- [00:00:04.430]and today I will be presenting on
- [00:00:06.163]"Improving the Efficacy of the
- [00:00:07.720]Tracer Algorithm as a Gap-Filling
- [00:00:09.385]and Pathway Prediction Tool
- [00:00:10.669]for Metabolic Models."
- [00:00:18.199]Genome-scale metabolic models are
- [00:00:20.098]important tools
- [00:00:20.959]that link an organism's genes,
- [00:00:22.508]proteins, and reactions in
- [00:00:24.118]reconstructions of their metabolic
- [00:00:25.856]network. These models can predict cellular
- [00:00:28.116]functions by simulating the flow
- [00:00:29.703]of compounds, or metabolites,
- [00:00:31.520]through this network.
- [00:00:33.420]However, missing reactions
- [00:00:34.850]known as "gaps" in these
- [00:00:36.098]reconstructions lead to dead-end,
- [00:00:38.190]or blocked, metabolites
- [00:00:39.384]and hence limits application of the model.
- [00:00:43.487]The Tracer algorithm is a novel
- [00:00:45.175]gap-filling tool that uses
- [00:00:46.509]a brute-force, confidence-based approach to solve blockages.
- [00:00:49.858]The tool attempts to enable blocked pathways
- [00:00:51.730]using reactions from
- [00:00:52.615]other species contained in a set of
- [00:00:53.737]databases.
- [00:00:55.646]Once a viable pathway is found,
- [00:00:57.059]its confidence score is calculated
- [00:00:58.989]based on the origin of each reaction.
- [00:01:01.420]If the confidence score falls
- [00:01:02.592]within a user-determined threshold,
- [00:01:04.358]the pathway is passed as
- [00:01:05.561]a solution.
- [00:01:06.874]After enough unsuccessful attempts,
- [00:01:08.312]the algorithm attempts to
- [00:01:09.632]solve the blockage by gradually
- [00:01:12.209]relaxing parameters, a process
- [00:01:14.205]known as rule-breaking.
- [00:01:20.086]As the Tracer algorithm is a new
- [00:01:21.411]and developing tool, the
- [00:01:22.576]primary objectives of this
- [00:01:23.761]research included optimization
- [00:01:24.846]of the algorithm and validation
- [00:01:26.714]of its results.
- [00:01:31.073]To develop and test
- [00:01:32.520]the algorithm, a model of the
- [00:01:34.356]plant species, maize, was used.
- [00:01:36.476]The Present Database then
- [00:01:37.419]contained all the maize reactions,
- [00:01:39.100]the Super Database contained
- [00:01:40.361]reactions from closely-related plants,
- [00:01:42.435]the Ortholog Database contained orthologs
- [00:01:44.411](or genes with the same function in
- [00:01:46.293]different organisms) for each of the
- [00:01:48.151]species in the Super Database,
- [00:01:49.762]and the Master Database contained
- [00:01:51.088]all known plant reactions.
- [00:01:54.648]However, formatting
- [00:01:55.444]inconsistencies in the Super and
- [00:01:57.057]Ortholog databases prevented
- [00:01:58.740]the algorithm from assigning accurate
- [00:02:00.544]confidence scores according to
- [00:02:01.875]the gene-protein-reaction, or GPR,
- [00:02:04.198]association-based process
- [00:02:05.570]depicted in Figure 3.
- [00:02:08.303]To address the issue,
- [00:02:09.386]a standardized format was
- [00:02:10.615]applied to the GPR information
- [00:02:12.260]contained within the Super
- [00:02:13.071]Database to improve readability
- [00:02:15.482]and the algorithm was adapted
- [00:02:16.641]to read this new
- [00:02:17.466]format.
- [00:02:18.764]Additionally, a new
- [00:02:19.764]Ortholog Database was compiled
- [00:02:21.126]using the original, as well as five
- [00:02:23.462]open-source databases containing
- [00:02:25.415]the orthologs for each species.
- [00:02:28.161]The gene names were then
- [00:02:28.997]translated to match the
- [00:02:29.633]naming convention of the genes
- [00:02:31.583]in the Super Database.
- [00:02:34.487]By analyzing information
- [00:02:35.649]from each trial, it was
- [00:02:37.196]determined that the
- [00:02:37.846]algorithm was testing
- [00:02:38.553]reactions in a haphazard
- [00:02:39.806]manner.
- [00:02:40.960]The diversification of the pool of
- [00:02:42.478]possible reactions--intended to enable
- [00:02:44.919]a more thorough search once
- [00:02:46.052]the original pool had failed--was
- [00:02:47.995]being applied ineffectively and
- [00:02:49.383]reactions were being
- [00:02:50.165]tested in a random order.
- [00:02:52.378]Additionally,
- [00:02:53.009]the algorithm was
- [00:02:53.544]testing a large number
- [00:02:54.351]of pathways that had
- [00:02:55.790]previously failed due to delayed
- [00:02:57.511]rule-breaking and failure of a
- [00:02:58.604]metabolite, which
- [00:02:59.587]resulted in long run-times.
- [00:03:03.193]Finally, to begin validation
- [00:03:04.277]of the algorithm, the reaction
- [00:03:06.056]producing pelargonidin 3-glucoside
- [00:03:08.721]from pelargonidin--part of the
- [00:03:10.581]well-studied anthocyanin subsystem
- [00:03:12.562]found in many plant species--was
- [00:03:14.319]removed from the model. The algorithm
- [00:03:16.627]was then executed to
- [00:03:17.664]solve this blockage and the
- [00:03:19.004]result compared to the known pathway.
- [00:03:24.084]According to the previous analysis,
- [00:03:25.683]changes were made to the algorithm
- [00:03:26.955]to improve its functionality. In
- [00:03:29.337]the most recent version,
- [00:03:30.590]diversification of the reaction
- [00:03:32.382]pool occurred once the
- [00:03:33.688]original pool had been exhausted,
- [00:03:35.474]rather than on every even trial,
- [00:03:37.056]like the original algorithm.
- [00:03:39.155]Instead of randomly testing reactions
- [00:03:40.582]from the solution pool,
- [00:03:42.452]reactions were tested in order
- [00:03:43.589]from best to worst confidence score.
- [00:03:46.367]FInally, the algorithm, which originally
- [00:03:47.734]checked only every ten
- [00:03:48.973]trials to see if it had exhausted
- [00:03:50.412]all possibilities and should engage
- [00:03:52.180]rule-breaking or fail a metabolite early,
- [00:03:54.902]also began to check for exhaustion
- [00:03:56.064]of possibilities after the
- [00:03:57.448]reaction pool had been
- [00:03:58.047]diversified.
- [00:04:01.263]As shown here, the updated
- [00:04:02.636]algorithm was able to solve more
- [00:04:04.160]blocked metabolites and provide a GPR
- [00:04:05.939]verification for about
- [00:04:07.571]sixty-four percent of the
- [00:04:08.551]reactions it added, compared to
- [00:04:09.797]twenty-two percent
- [00:04:10.618]in the original algorithm. Finally
- [00:04:12.984]the updated algorithm used about
- [00:04:14.413]forty percent of the number
- [00:04:15.557]of trials as the original.
- [00:04:21.995]Finally, the algorithm was
- [00:04:23.058]able to successfully fill the gap
- [00:04:24.640]introduced in the anthocyanin
- [00:04:26.070]pathway as shown in the figure.
- [00:04:32.563]The ability of the Tracer algorithm
- [00:04:34.120]to fill the gap introduced
- [00:04:35.201]in the anthocyanin
- [00:04:36.201]pathway suggests that the
- [00:04:37.687]algorithm can be used to
- [00:04:38.971]assist with modelling of
- [00:04:40.069]subsystems typically overlooked
- [00:04:41.740]by other gap-filling algorithms.
- [00:04:44.434]The large path functionality of
- [00:04:45.355]the Tracer algorithm, which enables
- [00:04:47.044]the creation of extensive, sprawling
- [00:04:48.528]pathways, can also help with pathway
- [00:04:51.199]discovery by generating targets
- [00:04:52.683]for lab verification.
- [00:04:59.205]Before the Tracer algorithm can be
- [00:05:00.559]used extensively, additional
- [00:05:02.374]verification must be
- [00:05:03.615]conducted using more complex pathways.
- [00:05:06.865]The algorithm should also be
- [00:05:08.199]generalized to work with any
- [00:05:09.505]species--given the correct inputs--
- [00:05:11.486]and its user-interface and pipeline
- [00:05:12.999]should be improved to facilitate
- [00:05:14.382]this process.
- [00:05:18.810]This work was funded by the National
- [00:05:20.135]Science Foundation, and I would like
- [00:05:21.853]to thank them, and the University
- [00:05:23.630]of Nebraska - Lincoln, Center for Root
- [00:05:25.608]and Rhizobiome Innovation,
- [00:05:27.418]for their support.
- [00:05:28.765]Thank you.
The screen size you are trying to search captions on is too small!
You can always jump over to MediaHub and check it out there.
Log in to post comments
Embed
Copy the following code into your page
HTML
<div style="padding-top: 56.25%; overflow: hidden; position:relative; -webkit-box-flex: 1; flex-grow: 1;"> <iframe style="bottom: 0; left: 0; position: absolute; right: 0; top: 0; border: 0; height: 100%; width: 100%;" src="https://mediahub.unl.edu/media/13790?format=iframe&autoplay=0" title="Video Player: Improving the Efficacy of the Tracer Algorithm as a Gap-Filling and Pathway Prediction Tool for Metabolic Models" allowfullscreen ></iframe> </div>
Comments
0 Comments