A Reactome-guided map of inflammation, barrier failure, and colitis-associated colorectal cancer
A gene list becomes biologically useful when it is reorganized into processes: immune activation, immune restraint, epithelial cohesion, damage sensing, and escape from cell-cycle control. Reactome helps with that reorganization because it represents curated reactions and pathway hierarchies rather than associations inferred from co-expression alone (Milacic, Beavers, Conley, and others, 2024). But it also exposes an important limitation: several of the most interesting methylation candidates are poorly characterized and do not yet have direct curated Reactome pathways.
The Reactome mappings were queried on 25 August 2026 with HGNC symbols, projection to human, and interactors disabled. The disease-stage scores below are qualitative hypotheses synthesized from the cited tissue literature; they are not patient measurements, differential-expression results, or pathway-enrichment p-values.
The panel combines immune genes that were reported as hypomethylated in severe UC (CD86, IL10, NLRC4, NLRP3, SIGLEC5, CLMP) with promoter-hypermethylated or neoplasia-marker candidates (FAM217B, KIAA1614, RIBC2, FOXE1, SYNE1) and established epithelial/cancer-relevant genes (CDH1, CDKN2A, GDNF) (Kang, Bae, Han, Kim, Kim, and Yi, 2016; Papadia, Louwagie, Del Rio, Grooteclaes, Coruzzi, Montana, Novelli, Bordi, de’ Angelis, Bassett, Bigley, Warren, Atkin, and Forbes, 2014; Saito, Kato, Hiraoka, Horii, Suzuki, Higashi, Kaji, Kondo, and Yamamoto, 2011). These are not interchangeable signals: some are pathway components, some are tissue markers, and some remain mechanistically under-annotated.
Direct queries returned curated neighborhoods for nine genes. Examples include CD28/CTLA-4 and IL-10 signaling for CD86; CDH1 transcription, translation, and adherens-junction organization; oncogene- and oxidative-stress-induced senescence for CDKN2A; RET/NCAM signaling for GDNF; the IPAF inflammasome for NLRC4; the NLRP3 inflammasome; and neutrophil degranulation for SIGLEC5. SYNE1 mapped to meiotic synapsis, a real annotation but not an obvious colon mechanism. CLMP, FAM217B, FOXE1, KIAA1614, and RIBC2 were not directly mapped by this identifier analysis.
Figure 1 makes that asymmetry visible. “Limited” does not mean unimportant: it means the database does not support a direct reaction-level bridge for the submitted symbol under these settings.
fig, ax = plt.subplots(figsize=(10,7))
sns.heatmap(M, cmap=sns.color_palette(["#F7FAFC",teal],as_cmap=True), cbar=False,
linewidths=1,linecolor="white",square=False,ax=ax)
ax.set(xlabel="Pathway neighborhood",ylabel="Candidate gene")
ax.tick_params(axis="x",rotation=0); ax.tick_params(axis="y",rotation=0)
fig.tight_layout(); fig.savefig(OUT/"reactome-coverage.png",dpi=220,bbox_inches="tight"); plt.show()
The central insight is not a single enriched pathway. It is a set of bridges between otherwise distinct programs, shown in Figure 2:
fig, ax = plt.subplots(figsize=(12,8)); ax.set(xlim=(-6,6),ylim=(-4.5,4.5)); ax.axis("off")
mod_pos={"Immune regulation":(-3.7,2.7),"Inflammasomes":(0,3.2),"Barrier / adhesion":(3.7,2.7),
"Trophic signaling":(-3.7,-2.5),"Senescence":(0,-3.1),"Under-annotated":(3.7,-2.5)}
gene_groups={"Immune regulation":["IL10","CD86","SIGLEC5"],"Inflammasomes":["NLRP3","NLRC4"],
"Barrier / adhesion":["CDH1","CLMP"],"Trophic signaling":["GDNF"],"Senescence":["CDKN2A"],
"Under-annotated":["FAM217B","KIAA1614","RIBC2","FOXE1","SYNE1"]}
gene_pos={}
for mod,(mx,my) in mod_pos.items():
ax.add_patch(FancyBboxPatch((mx-1.3,my-.42),2.6,.84,boxstyle="round,pad=.08",fc=navy,ec="none"))
ax.text(mx,my,mod,ha="center",va="center",color="white",weight="bold")
gs=gene_groups[mod]
span = 2.0 if mod=="Under-annotated" else 1.0
offsets=np.linspace(-span,span,len(gs)) if len(gs)>1 else [0]
for off,g in zip(offsets,gs):
gx=mx+off; gy=my-1.15 if my>0 else my+1.15; gene_pos[g]=(gx,gy)
col=gray if mod=="Under-annotated" else teal
width = .92 if mod=="Under-annotated" else 1.1
ax.add_patch(FancyBboxPatch((gx-width/2,gy-.25),width,.5,boxstyle="round,pad=.05",fc=col,ec="white"))
ax.text(gx,gy,g,ha="center",va="center",color="white",weight="bold",fontsize=8 if mod=="Under-annotated" else 9)
ax.plot([mx,gx],[my-.45 if my>0 else my+.45,gy+.27 if my>0 else gy-.27],color=col,lw=1.8)
bridges=[("Immune regulation","Inflammasomes","immune restraint / activation"),
("Inflammasomes","Barrier / adhesion","pyroptosis · permeability"),
("Barrier / adhesion","Senescence","injury · repair · selection"),
("Senescence","Under-annotated","methylation markers"),
("Trophic signaling","Barrier / adhesion","survival · tissue repair")]
for a,b,label in bridges:
x1,y1=mod_pos[a]; x2,y2=mod_pos[b]
ax.add_patch(FancyArrowPatch((x1,y1),(x2,y2),arrowstyle="-|>",mutation_scale=14,
linestyle="--",color=coral,lw=1.8,connectionstyle="arc3,rad=.08",shrinkA=55,shrinkB=55))
ax.text((x1+x2)/2,(y1+y2)/2+.18,label,ha="center",fontsize=8,color=coral,
bbox=dict(fc="white",ec="none",alpha=.8,pad=1))
ax.text(0,4.15,"Curated neighborhoods connected by disease-level hypotheses",ha="center",weight="bold",fontsize=14,color=navy)
fig.tight_layout(); fig.savefig(OUT/"reactome-network.png",dpi=220,bbox_inches="tight"); plt.show()
The pathway signal expected in a biopsy depends on both regulation and cell composition. Active colitis should most strongly expose innate sensing, leukocyte co-regulation, epithelial junction disruption, and compensatory IL-10. Long-standing colitis can add senescence and methylation-field signals. Established cancer may retain inflammatory circuitry while making cell-cycle escape and epithelial plasticity more prominent. Figure 3 is a qualitative expectation map, not a measured heatmap.
stage = pd.DataFrame({
"Healthy colon":[1,0,3,0,2,0],
"Active colitis":[3,3,1,2,2,1],
"Long-standing UC":[2,2,1,3,1,2],
"Colitis-associated CRC":[2,1,0,3,1,3]}, index=["Immune co-regulation","Inflammasome activity","Barrier integrity","Senescence / escape","Trophic signaling","Candidate methylation markers"])
fig, ax = plt.subplots(figsize=(9,5.5))
sns.heatmap(stage,cmap=sns.light_palette(coral,as_cmap=True),vmin=0,vmax=3,annot=True,fmt="d",linewidths=1,linecolor="white",cbar_kws={"label":"Hypothesized prominence"},ax=ax)
ax.set(xlabel="Disease context",ylabel="Pathway program")
ax.tick_params(axis="x",rotation=18); ax.tick_params(axis="y",rotation=0)
fig.tight_layout(); fig.savefig(OUT/"reactome-stage-map.png",dpi=220,bbox_inches="tight"); plt.show()
| Gene | Direct Reactome neighborhood from the query | Colitis/CRC interpretation |
|---|---|---|
| CD86 | IL-10 signaling; CD28 Vav1 and PI3K/AKT; CTLA-4 co-inhibition | APC abundance and immune context; not a unidirectional activation marker |
| CDH1 | CDH1 transcription/translation; type-I cadherins; adherens junctions | Barrier architecture, epithelial plasticity, and methylation-associated silencing |
| CDKN2A | Oncogene- and oxidative-stress-induced senescence | Stress arrest in colitis; loss can permit neoplastic escape |
| CLMP | No direct mapping in this query | Strong developmental intestinal genetics, uncertain adult UC pathway (Rathjen and Jüttner, 2023) |
| FAM217B | No direct mapping | UC methylation candidate; mechanism unknown (Kang, Bae, Han, Kim, Kim, and Yi, 2016) |
| FOXE1 | No direct mapping | Neoplasia-associated methylation marker; colon driver role unproven (Papadia, Louwagie, Del Rio, Grooteclaes, Coruzzi, Montana, Novelli, Bordi, de’ Angelis, Bassett, Bigley, Warren, Atkin, and Forbes, 2014) |
| GDNF | RET signaling; NCAM1 interactions and neurite outgrowth | Trophic stromal/neural signaling; inflammation-associated methylation (Saito, Kato, Hiraoka, Horii, Suzuki, Higashi, Kaji, Kondo, and Yamamoto, 2011) |
| IL10 | IL-10 production/signaling neighborhoods | Compensatory immune brake; ligand abundance does not prove response (Glocker, Kotlarz, Boztug, and others, 2009) |
| KIAA1614 | No direct mapping | UC promoter-methylation candidate with predicted, not established, polarity biology |
| NLRC4 | IPAF/NLRC4 inflammasome | Bacterial invasion sensing after barrier breach |
| NLRP3 | NLRP3 inflammasome | DAMP and organelle-stress sensing; activation requires more than RNA or methylation |
| RIBC2 | No direct mapping | Cilia biology in Xenopus; colon relevance unresolved (Chung, Kwon, Tu, Brooks, Gupta, Meyer, Baker, Marcotte, and Wallingford, 2014) |
| SIGLEC5 | Neutrophil degranulation; immunoregulatory interactions | Myeloid abundance and inhibitory/activating paired-Siglec balance |
| SYNE1 | Meiotic synapsis in this query | The colon signal is more defensible as a methylation biomarker than this pathway mapping |
The graph suggests measuring pathway outputs rather than adding more isolated genes. In paired tissue and plasma, a compact follow-up panel could combine:
That design keeps three questions separate: Which cells are present? Which pathways are active? Which methylation changes add predictive information? Reactome is strongest for the second question. It cannot, by itself, convert an under-characterized methylation marker into a mechanism.