With millions of cases of bladder cancer (BC) around the world, the need for tools to ensure a timely diagnosis of this condition is a matter of concern. Scientists recently used mitochondrial-related genes (MRGs), known to be involved in the progression of the disease, to build a novel diagnostic model using machine learning (ML).
The results were published in Scientific Reports, indicating the potential of this model pending further validation.
Study: Machine-learning prediction of a novel diagnostic model using mitochondria-related genes for patients with bladder cancer. Image Credit: mi_viri/Shutterstock.com
Bladder cancer
Bladder cancer is three to four times more common in men than in women, making it the sixth leading cause of cancer among men. It is primarily caused by smoking and exposure to certain industrial chemicals and typically affects middle-aged and older men.
Although bladder cancer is more prevalent in developed populations, its prognosis remains relatively poor despite medical advancements. This has driven the development of better diagnostic tools, prognostic models, and therapeutic approaches.
Mitochondria, the subcellular organelles responsible for energy production, control cell metabolism and regulate key cellular processes like programmed cell death, signaling, and calcium ion levels.
Tumor cells, which require a lot of energy, predominantly use glycolysis—a less efficient anaerobic pathway—unlike normal cells that rely on oxidative phosphorylation, a more efficient aerobic pathway that can produce up to 15 times more energy.
This difference in energy production is part of the ‘Warburg effect,’ where abnormal mitochondrial function leads to altered metabolism in tumor cells. For instance, malfunctioning mitochondria may prevent cancer cells from undergoing programmed death, allowing them to survive and spread.
Additionally, mitochondrial abnormalities can cause oxidative stress on cellular components such as DNA and proteins, increasing cancer risk, conferring resistance to cancer therapies, and promoting tumor growth.
“Given the important roles of MRGs in BC progression, it is important to screen novel biomarkers based on MRGs for BC patients.”
ML forms part of the artificial intelligence (AI) arsenal, which identifies patterns and knowledge from raw data without feeding in detailed instructions.
This allows the system to predict, categorize, and recognize trends that could include tumor-associated transcription patterns. In the current study, the researchers sought to exploit the power of ML on transcriptomes to build a new diagnostic model for BC based on MRGs.
What did the study show?
The researchers analyzed 165 bladder cancer (BC) samples and 67 controls to study the differential expression of mitochondrial-related genes (MRGs). They identified 752 differentially expressed MRGs, with 440 showing increased expression and the rest downregulated.
These genes were significantly involved in cell pathways related to organ formation in embryos, cell fate, transcription regulation, neurodegenerative diseases, and muscle tissue disorders.
The analysis identified nearly 50 BC-related features and narrowed down to 13 critical genes. Among these, TRAF3 Interacting Protein 3 (TRAF3IP3), Oxidative Stress-Induced Growth Inhibitor Mitochondrial (OXSM), N-myristoyltransferase 1 (NMT1), and Glutaredoxin 2 (GLRX2) were found to be key targets. GLRX2, in particular, is important for maintaining the oxidation-reduction balance within mitochondria, which helps normal cellular processes to continue without oxidative damage.
Expression patterns of GLRX2, NMT1, OXSM, and TRAF3IP3 showed clear differences between BC samples and controls, achieving 90% effectiveness in differentiation. GLRX2, NMT1, and OXSM were highly upregulated in BC, whereas TRAF3IP3 was significantly reduced.
These findings were consistent across two additional datasets, demonstrating that this model more effectively differentiates BC from control samples than single gene biomarkers.
Furthermore, the study explored where these genes were predominantly expressed, finding that different pathways and immune cells in the tumor microenvironment responded variably to changes in gene regulation. For instance, higher levels of activated natural killer (NK) and plasma cells were associated with increased GLRX2 expression.
NMT1 expression, which was significantly raised in several BC cell lines, encodes a protein crucial for protein modification and signaling, potentially enhancing tumor cell interactions with the extracellular matrix—a key process in cancer spread. Significantly, suppressing NMT1 led to restrained BC cell growth, indicating its role in advancing BC.
Conclusions
The advent of transcriptomics and ML in tumor diagnostic models has brought impetus to the search for the accurate and early diagnosis of BC without the need for invasive and painful biopsies. This ML approach can help to frame personalized diagnostic and treatment plans based on biomarker selection.
It can also speed up decisions with increased efficiency. Finally, it can help understand the process of tumor development via the insights it provides into the underlying biology of the tumor.
The current study identified four genes (GLRX2, NMT1, OXSM and TRAF3IP3) for BC diagnosis. These were incorporated into a diagnostic model. They were also found to play important roles in BC progression. Further research is essential to confirm these findings in a more diverse sample.
“Our findings could potentially lead to enhanced accuracy and reliability in diagnosing BC, contributing to more personalized and effective medical interventions for patients in the future.”
