- by Torie Noellsch
Artificial intelligence is helping Texas A&M AgriLife Research scientists fast-track answers to a long-standing question in human biology: What do hundreds of mystery proteins in the mitochondria actually do?
For years, scientists have known that about 1,100 proteins exist within mitochondria, the energy centers of human cells, but they still don't know what many of them do.
This presents a problem because mutations in nearly 30% of mitochondrial proteins have been linked to human diseases, from more common conditions like cancer, Parkinson's and Alzheimer's to rare inherited metabolic disorders.
Abhinav Swaminathan, Ph.D., and Vishal Gohil, Ph.D., Texas A&M AgriLife Research scientists in the Texas A&M Department of Biochemistry and Biophysics, perform laboratory experiments to validate the mitochondrial protein pairings predicted by the AI tool AlphaFold-Multimer. (Michael Miller/Texas A&M AgriLife)
Pairing up proteins for answers
Without knowing the precise biochemical function of these proteins, it is difficult to understand the causes of these diseases. One way to infer a protein's function is to identify which other proteins it interacts with.
In the study " The predicted interactome of the human mitochondrial proteome ," published in Nature Communications , the team used an AI tool to screen every possible pairing of about 1,100 human mitochondrial proteins — more than 630,000 combinations — for likely interactions. The predictions linked many disease-associated proteins to previously unknown biological pathways, pointing to potential diagnostic and therapeutic targets.
From a basic science perspective, the study also gave 85 previously uncharacterized proteins their first probable partners, offering clues about what those proteins may do in cells.
"If we don't know the function of these individual proteins, we cannot determine how to treat the diseases they may cause," said Abhinav Swaminathan, Ph.D., the study's lead author and an AgriLife Research scientist in the lab of Vishal Gohil , Ph.D., professor and Chancellor EDGES Fellow in the Texas A&M Department of Biochemistry and Biophysics .
Repurposing an AI tool that predicts protein shape
Researchers nationwide have been working to identify the functions of the mystery mitochondrial proteins to solve this problem, but traditional lab methods are slow.
Research efforts led by Abhinav Swaminathan, Ph.D., and Vishal Gohil, Ph.D., Texas A&M AgriLife Research scientists in the Texas A&M Department of Biochemistry and Biophysics, used AI tool AlphaFold-Multimer to reveal 2,895 likely mitochondrial protein interactions. (Michael Miller/Texas A&M AgriLife)
Nearly three years ago, Swaminathan and Gohil set out to speed things up by testing an artificial intelligence tool called AlphaFold-Multimer, or AFM.
Google DeepMind created and unveiled AFM in 2022 to predict the three-dimensional shapes that proteins form together.
"We didn't have reliable experimental tools to evaluate direct mitochondrial protein interactions for the mitochondrial copper pathway proteins we study, so we ran a trial using the AI approach," Swaminathan said. "It worked really well, so we decided to apply it to all human mitochondrial proteins."
Out of the 630,003 possible protein pairs, the team cataloged 2,895 likely interactions. More than half of them had never been reported before.
Confirming AI predictions in the lab
About two years of laboratory experiments followed, confirming several of the predicted partnerships in yeast and human cells. Some of these experiments were performed in collaboration with researchers at Washington University School of Medicine in St. Louis.
The results showed AI predictions can serve as a road map, pointing scientists toward lab experiments most likely to yield results. The team made its predictions available as a free online resource, MitoMatch , so researchers worldwide can assign likely partners to the proteins they study.
"I think this research was successful because we asked the bigger question about what we could do to address the protein function problem, and we got creative with our resources to develop a solution that anyone can use."
Abhinav Swaminathan, Ph.D.
Texas A&M AgriLife Research
Department of Biochemistry and Biophysics
As team members continue experimenting with the data, they are also collaborating with other researchers to better understand cell function and work toward new insights into human disease.
Cures for mitochondria-linked diseases may still be far into the future, but Swaminathan said the team's recent success with its AI approach is a promising starting point.
"We are a traditional biochemistry and genetics lab, not an AI lab," Swaminathan said, "but I think this research was successful because we asked the bigger question about what we could do to address the protein function problem, and we got creative with our resources to develop a solution that anyone can use."
This study was supported by the National Institutes of Health , Texas A&M High Performance Research Computing and The Welch Foundation .
