How the Kozik Lab Uses SeqHub to Make Sense of the Understudied Respiratory Microbiome
Ariangela Kozik is an assistant professor at the University of Michigan, jointly appointed in the Department of Molecular, Cellular, and Developmental Biology and the Department of Internal Medicine. Her lab studies host-microbe and microbe-microbe interactions in the human respiratory system, with a focus on how microbial communities contribute to chronic respiratory disease.
Much of that work involves organisms that are common in the airway but poorly characterized. The lab works from bacterial isolates cultured directly from patient samples, often full genomes, and regularly runs into genes with no confirmed function and little representation in reference databases.
Before SeqHub
The lab's workflow relied on BLAST for sequence homology, but annotation quality was inconsistent, pinpointing the exact genomic location of a gene was often difficult, and retrieving the genomic neighborhood around a gene of interest was its own multi-step process. Structural work meant separate tools, PyMOL and AlphaFold, with their own learning curves and requiring comfort with the command line, which is not a given across a mixed wet and dry lab.
As a result, hypothetical proteins were usually where the process stalled, an issue Dr. Kozik has run into since her postdoc.
Hypothetical proteins were the end of the road. It would take a combination of databases and tools, and it wasn't something that seemed worth the time.
Ariangela Kozik, University of Michigan
What changed with SeqHub
For the Kozik Lab, SeqHub replaced a patchwork of separate tools with a single place to work from a sequence. That consolidation, and the ease of access it provides to structure, sequence, annotation, genomic context, and more, surfaces connections that piecing together separate tools would not have made visible.
Now we can use the context-based information in SeqHub, and even that on its own is really helpful. But then also being able to see where similar proteins exist in other bacteria…[SeqHub] provides more peripheral information that can be useful, whereas before you just had a sequence that you'd BLAST and hope for the best. That has really helped make more inference possible that wasn't possible before.
Ariangela Kozik, University of Michigan
PPI prediction, HMM domain search, and co-occurrence search (CoSearch) are also part of that same picture, available alongside annotation and genomic context rather than requiring separate tools or comfort with the command line. (An API is also available for labs that want to work that way.)
That shift shows up across the lab in a few distinct ways.
How the lab uses it
Use case 1: Finding transposons and other domains
Kelyah Spurgeon works with a QAC protein (a transporter associated with resistance to quaternary ammonium compounds, common disinfectants) and uses SeqHub's tools to identify transposons and other domains around the gene that were not previously obvious.
“From SeqHub I can see the structure, amino acid sequence, annotation, and the locus for my Qac protein. On top of that, the AI agent can list relevant evidence towards potential function for me and can identify whether my proteins of interest potentially exist in an operon. With other annotation software there are just fewer descriptors. With SeqHub I'm getting annotation and sequence information directly from my genome of interest and it makes it easier to investigate.”
— Kelyah Spurgeon, Kozik Lab
Use case 2: Surfacing candidate genes literature alone would not show
Claire Albright used SeqHub's PPI network view while working from a list of candidate genes tied to a specific response system, to identify new candidates to test.
“Looking at the predicted protein-protein interactions was extremely helpful. Based on the literature, I had the functions [of some genes]…but SeqHub also allowed me to identify other potential functions within that network, and cross-reference that with the genomic context to see if there were any closely related genes predicted to have similar function…I could add them to the list of genes of interest. This is a quick and comprehensive way of getting at a question that would have taken a lot of time if I were doing it individually on AlphaFold.”
— Claire Albright, Kozik Lab
Use case 3: Understanding pathways by layering genomic context with domains
Sophie Pickelner was working to understand which genes are involved in steroid metabolism, and used SeqHub to layer HMM domain search with genomic context across species.
“I was looking for genes predicted to be involved in steroid metabolism and there weren't many annotated. But looking through HMM domains alongside related proteins across species that are involved in steroid metabolism, I was able to find some candidate annotations.”
— Sophie Pickelner, Kozik Lab
From prediction to the bench
In one case, whole-genome sequencing of a human isolate turned up a disinfectant resistance gene that is not typically present in that organism. Other protein databases were not particularly helpful because the genes were named inconsistently across sources.
We put this [isolate] through SeqHub and got higher resolution information about the predicted proteins…and more confidence to say, yes, this particular protein, despite the inconsistent naming across databases and annotations, seems to match a family of proteins that are well characterized in other organisms. That increased resolution gave us some direction. We then were able to take those predictions back to the wet lab to begin to experimentally validate the function [from the SeqHub predicted annotation].
Ariangela Kozik, University of Michigan
Closing the annotation gap
For labs working with understudied or patient-derived microbes, annotation gaps are the norm. The Kozik Lab's work shows what's possible when annotations, genomic context, PPI networks, and HMM domain search are available from the start, from spotting domains and candidate genes other tools miss to following a disinfectant resistance gene from sequence to the bench.
“SeqHub is a great platform for genome gazing. It pulls together several sources of information, integrates the unique genomic neighborhood perspective, and really is a great starting point when you have a new genome or protein that you want to study. Staying up to date with new tools can feel daunting when there are a million other demands on your time, but I highly recommend booking an info session for your team/lab group.”
— Ariangela Kozik, University of Michigan
SeqHub is free for academic use
Start annotating your sequences
Join research labs using SeqHub to draw more inference from the same genomic context, PPI networks, and domain signals already sitting in your data.