Cracking the Code of UC: How Lactylation, AI, and Docking Reveal New Biomarkers and Treatments

Hey there! Let’s chat about something pretty significant in the world of health, specifically a tricky condition called Ulcerative Colitis, or UC for short. If you or someone you know deals with it, you know it’s no walk in the park. It’s a chronic inflammatory bowel disease that can really mess things up, causing all sorts of unpleasant symptoms like abdominal pain, weight loss, and, well, bloody diarrhea. Not fun.

UC has been on the rise globally, and while we’ve learned a lot, finding reliable ways to diagnose it early and treat it effectively is still a big challenge. Current therapies often aren’t enough, and the disease is just so different from person to person. That’s where the hunt for new clues comes in.

Now, here’s where it gets interesting. Scientists have been looking into something called lactylation. Think of it as a tiny chemical tag that gets added to proteins. This tag, derived from lactate (yes, like the stuff that builds up in your muscles after a tough workout!), can change how proteins work and affect things like metabolism and immune responses. Recent studies hinted that lactylation might be playing a sneaky role in gut inflammation, maybe even through how our genes are regulated. But the exact details in UC? Still a bit fuzzy.

So, we thought, “Okay, how can we really dig into this lactylation thing in UC and see if it holds any answers?” And that’s where the cool tools come in: machine learning and molecular docking. These aren’t just fancy buzzwords; they’re powerful ways to sift through massive amounts of data and predict how molecules might interact.

How We Tackled This Mystery

Our mission was clear: find out if lactylation-related genes are acting differently in people with UC compared to healthy folks, see if any of these genes could be used as biomarkers for diagnosis, understand *how* they might be involved, and maybe, just maybe, find existing drugs that could target them.

We started by grabbing some publicly available data – think of it as huge spreadsheets containing information about gene activity from lots of UC patients and healthy individuals. We combined a couple of these datasets to get a stronger starting point, making sure to smooth out any differences between them (like batch effects – imagine trying to compare notes taken with different pens on different paper; we made them look uniform!).

Next, we looked for genes that were acting significantly differently (either much higher or much lower activity) in the UC samples compared to the normal ones. We also got a list of genes known to be involved in lactylation from previous studies. The sweet spot? The genes that were *both* acting differently in UC *and* related to lactylation. We found 22 of these!

To figure out what these 22 genes might be doing, we did some functional analysis. It’s like asking, “Okay, you 22 genes, what jobs are you usually involved in?” Turns out, they were heavily linked to things like actin cytoskeleton organization (think of the cell’s internal scaffolding) and the JAK-STAT signaling pathway (a key communication route for immune cells). This immediately gave us clues that lactylation in UC might be messing with cell structure and immune signals.

Meet the Biomarkers

Finding 22 genes is great, but which ones are the *most* important? This is where machine learning shines. We fed the data on these 22 genes into three different machine learning algorithms: LASSO, Random Forest (RF), and SVM-RFE. Each one has a slightly different way of picking out the most important players.

LASSO suggested 9 genes, SVM found 17, and RF pointed to 4. But we were looking for the *real* VIPs – the genes that all three methods agreed were super important. And guess what? Three genes popped up in the intersection of all three lists:

• S100A11
• IFI16
• HSDL2

These three became our “model genes” – the prime candidates for being diagnostic biomarkers. We checked their expression levels in the data, and indeed, S100A11 and IFI16 were significantly *upregulated* (more active) in UC patients, while HSDL2 was significantly *downregulated* (less active). This pattern was consistent across the different datasets we used for training and testing our models.

But could they *diagnose* UC? We built a diagnostic model using these three genes and tested its ability to distinguish between UC and healthy samples. The results were pretty impressive! Using something called an ROC curve, we measured the model’s performance. A perfect score is 1.0, and our three-gene model hit an AUC (Area Under the Curve) of 0.973 in our training data. We then tested it on a completely separate set of data (the validation set), and it still performed remarkably well. This told us these three genes, together, have excellent potential as biomarkers for diagnosing UC.

To double-check, we even did some experiments in the lab using human intestinal cell lines (Caco-2 and HT-29). We stimulated these cells to mimic the inflammation seen in UC and then measured the activity of our three model genes. The results mirrored what we saw in the patient data: S100A11 and IFI16 went up, and HSDL2 went down. It’s always nice when the lab work backs up the computer analysis!

Unpacking the ‘How’: Metabolism and Immunity

Okay, so we found potential biomarkers. But *how* are they involved in UC? We dug deeper into their possible functions, looking at metabolic pathways and the immune microenvironment.

Using another analysis method (GSVA), we saw that high levels of S100A11 and IFI16 were linked to pathways involved in breaking down certain amino acids (like valine, leucine, and isoleucine – the branched-chain amino acids, or BCAAs) and pyruvate metabolism. On the flip side, low levels of HSDL2 were associated with different metabolic processes, including amino acid synthesis and degradation.

This is a big deal because altered amino acid metabolism is already known to be connected to UC. Patients often have lower levels of BCAAs and changes in other amino acids like histidine. Our findings suggest that S100A11, IFI16, and HSDL2 might be influencing UC by messing with these metabolic pathways.

We also looked at how these genes correlated with different types of immune cells in the gut lining. This is crucial because UC is fundamentally an immune-driven disease. We found that S100A11 and IFI16 showed positive correlations with immune cells like neutrophils and certain types of macrophages (M0 and M1), which are often pro-inflammatory. Interestingly, they were negatively correlated with M2 macrophages, which are more involved in resolving inflammation and repair. HSDL2 showed the opposite pattern – negative correlations with neutrophils and M0 macrophages, and a positive correlation with M2 macrophages.

Think of macrophages like little clean-up and defense crews in your tissues. M1 types are the aggressive fighters, while M2 types are the peacekeepers and repair crew. In UC, this balance is often tipped towards the M1 pro-inflammatory state. Our results suggest that S100A11 and IFI16 might be pushing things towards inflammation by influencing these immune cells, while HSDL2 might be trying to pump the brakes.

We also peeked into which transcription factors (TFs – proteins that control gene activity) might be regulating our three model genes. We found two common TFs, CTCF and STAT1, that could be pulling the strings behind S100A11, IFI16, and HSDL2. Knowing the regulators helps us understand the whole network.

And just for good measure, we checked where these proteins hang out in the cell and in different cell types in the colon. S100A11 was found in the nucleus and cytoplasm, IFI16 also in the nucleus and cytoplasm, but HSDL2 was specifically in the mitochondria (the cell’s powerhouses). At the single-cell level in the colon, S100A11 showed up in Paneth cells and distal enterocytes, IFI16 in T cells, and HSDL2 in undifferentiated and enteroendocrine cells. This cellular zip code information gives us more clues about their specific roles.

Finding Potential Allies: The Drug Hunt

Okay, biomarkers are great for diagnosis, but what about treatment? Could we find existing drugs that might target the pathways or genes involved?

We used a database called CMap, which links gene expression patterns to drugs. We fed the gene signature of UC into CMap to see which drugs might reverse that pattern. This gave us a list of potential small molecule drugs.

Then came the molecular docking part. This is like a virtual “lock and key” experiment. We took the 3D structures of our model genes (the “locks”) and the structures of the top candidate drugs from CMap (the “keys”) and used software to see how well they fit together and how strongly they might bind. The better the fit and stronger the binding, the more likely the drug is to interact with the protein.

Out of the top candidates, two drugs showed the best docking scores with our three model genes: regorafenib and R-428. Their binding affinities were really strong (indicated by low docking energy scores, less than -9.0 kcal/mol – basically, a very snug fit!).

Now, these drugs aren’t currently standard treatments for UC. Regorafenib is a tyrosine kinase inhibitor used for certain cancers (like colorectal cancer), and R-428 is also a tyrosine-protein kinase receptor inhibitor being studied for various conditions. Their known mechanisms involve regulating cell growth, signaling, and blood vessel formation, which *could* potentially impact the inflammation and immune response seen in UC. While this is just a prediction based on computer modeling, it’s a really exciting lead for future research!

What’s Next? (And What We Couldn’t Do Yet)

This study gives us a fantastic starting point. We’ve identified three promising lactylation-related genes as potential biomarkers for UC diagnosis and pinpointed two existing drugs that *might* work as treatments by targeting these pathways. It really highlights that lactylation is likely playing a significant, previously underappreciated, role in UC.

However, science is a journey, not a destination! Our study, while thorough in its computational analysis, has limitations. We validated gene expression in cell lines, but we didn’t do functional experiments to prove *exactly* how S100A11, IFI16, and HSDL2 contribute to UC development or how regorafenib and R-428 might work. That’s the crucial next step – getting into the lab and doing experiments in cells and animal models to confirm these findings and understand the precise molecular mechanisms.

Also, our study looked back at existing data, and the sample sizes weren’t massive. To be absolutely sure about the diagnostic value of these biomarkers and the potential of these drugs, we need larger studies with more patients, following them over time.

Wrapping It Up

Despite the limitations, what we’ve done here is open a new door. By combining the emerging understanding of lactylation with the power of machine learning and molecular docking, we’ve proposed a novel strategy for tackling UC. Identifying S100A11, IFI16, and HSDL2 as potential diagnostic biomarkers and suggesting regorafenib and R-428 as possible therapeutic candidates provides a strong foundation for developing more targeted and effective ways to diagnose and treat Ulcerative Colitis in the future. It’s a big step forward in understanding this complex disease!

Source: Springer