Personally I think that both the common pathway databases and common pathway tools such as GSEA/GSVA/ORA-like-tools are all pretty much a mess. Pathway annotations are generic, highly redundant and either overly granular or far too inflated to be informative for your question. Likewise, tools like GSVA integrate dozens or hundreds of genes into a single "blackbox" score, so even in case of significance, you don't know if your treatments really have a specific effect, or whether it's the same genes (that might be redundant and unspecific to a certain pathway) driving the significance. A common example I like to cite is riboprotein genes, which e.g. in REACTOME are in like every 2nd pathway. These tend to give a lot of siginficant pathway results if included in your DE list / GSEA ranking, but obviously it's anything but specific.

What I would do is to center it all on your DE analysis. Do interaction tests, so test whether a certain treatment returns genes significnatly more different than other comparisons. Narrow down genes in particular effected in a certain treatment. Then, based on these genes check which are the central pathways they're affecting. ORA can help picking candidates, but don't rely on the stats. The hypergeometric test it uses is primitive and by far not suitable for complex data like that. In the end pathway analysis is a lot of manual curation as it's important to ensure that significant results are supported by key genes that the field accepts. A lot of modulator rather than driver genes might support a narrative, but it's hard to convice pathway people with it.

I’ve worked with similar comparisons, and I’d avoid relying only on overlapping enriched pathways because it can miss real differences. One approach worth trying is comparing pathway activity scores directly, like GSVA, ssGSEA, or PROGENy, where each sample gets a pathway-level score and you can test differences between drugs.

Another option is to run differential expression between drug conditions directly (Drug A vs Drug B) and then perform pathway analysis on those results. Methods like GSEA with ranked gene lists can also help identify pathways with stronger enrichment in one treatment.

I’d also check papers using pathway-level modeling, as they often handle these comparisons better than simple overlap methods.

Ingenuity Pathway Analysis (IPA) can help your. Comparison analysis of the differential expression results will let you look across all at once and find other drugs that look similar/dissimilar with Omicsoft analysis match/interpret, all in IPA.

In IPA specifically you can: Run a Core Analysis per drug-vs-control, then feed them into IPA's Comparison Analysis tool. It builds a heatmap (pathways × drugs) using activation z-scores and p-values, so instead of overlap counting you get quantitative, directional pathway activity per drug — letting you spot pathways strong in Drug A but flat in Drug B ("unique"), or same pathway with opposite/different-magnitude z-scores across drugs ("differentially enriched"). You can do the same at the Upstream Regulator level for mechanistic comparisons, and hierarchically cluster drugs by their pathway signatures.

You can download a trial here: https://digitalinsights.qiagen.com/products-overview/discovery-insights-portfolio/analysis-and-visualization/qiagen-ipa/?cmpid=ADMWEB

Good luck!