Page 68 - TD-4-1

Tumor Discovery Drug repurposing for pancreatic cancer via AI

Table 2. Filtering potential molecular drugs for pancreatic ductal adenocarcinoma based on three drug design specifications:
regulatory ability, sensitivity, and toxicity, derived from the candidate molecular drugs listed in Table 1
Target drug c‑MYC (+) FOXO3 (+) TP53 (*) Sensitivity (PRISM) Toxicity (LC50, mol/kg)
MK-2206 ↓ 0.772406631 5.561
Gemcitabine ↓  2.417963872 2.381
Notes:  denotes drug-target interaction; ↓ denotes downregulation of the biomarker by the molecular drug.
Abbreviations: FOXO3: Forkhead box O3; LC50: Lethal concentration 50%; PRISMA: Pharmaceutical Regulatory Information System; TP53: Tumor
suppressor p53.

where interactions between proteins and genes were 3.2. Core signaling pathways of carcinogenic
encoded as 1 (interaction) and 0 (no interaction). The mechanisms of PDAC
nodes in the GWGENs include proteins, genes, miRNAs, Tumors are more likely to accelerate their development in
and lncRNAs. an appropriate microenvironment. The microenvironment
Next, using systematic identification and order of PDAC involves complex interactions between tissues,
detection methods, false positives were removed from the blood vessels, immune cells, cytokines, and other molecules
39
candidate GWGENs based on whole-genome microarray surrounding pancreatic cancer cells. The characteristics
data GSE183795, resulting in the construction of the real of this microenvironment include fibrosis, immune cell
GWGEN for both PDAC and healthy controls, as shown infiltration, angiogenesis, and the presence of various
in Figure S1. Although the number of nodes was reduced cytokines and growth factors. 40,41 To identify relevant
after the false positives were filtered out from candidate signaling pathways based on their cellular functions,
GWGENs, the real GWGENs for both PDAC and healthy we consulted existing research literature. Key pathways
controls remained too complex for direct annotation using involved in pancreatic fibrosis include transforming growth
KEGG pathway analysis. To simplify this, the number factor β, Wnt, and phosphoinositide 3-kinase (PI3K)-
42
of nodes in GWGENs was reduced to 6,000 using the protein kinase B (AKT) signaling pathways. Pathways
PNP method. The 6,000 significant nodes from the real associated with immune cell infiltration include Janus
GWGENs of PDAC and healthy controls were extracted, kinase-signal transducer and activator of transcription
forming the core GWGENs for PDAC and healthy controls, (STAT) pathway, PI3K-AKT, and mammalian target of
43
as depicted in Figure S2. rapamycin (mTOR) signaling pathways. Additionally,
the mitogen-activated protein kinase, PI3K-AKT, and
For KEGG pathway enrichment analysis to interpret mTOR signaling pathways have been implicated in the
the carcinogenic mechanisms of PDAC, the DAVID creation of new blood vessels (angiogenesis). These six
44
functional annotation tool was employed, supported by signaling pathways provide a comprehensive framework
references from PDAC-related literature. Based on the for investigating the oncogenic mechanisms of PDAC.
KEGG pathway annotations, core signaling pathways for
PDAC and healthy controls were established, allowing Through KEGG enrichment pathway analysis and
the investigation of the mechanisms involved in PDAC annotation of the core GWGENs of PDAC and healthy
carcinogenesis, as illustrated in Figure 2. controls, we identified the core signaling pathways involved
in PDAC carcinogenesis, as shown in Figure 2. By examining
Based on these core signaling pathways, key biomarkers the core signaling pathways and their downstream target
of PDAC carcinogenesis were identified as potential drug genes, we explored the carcinogenic mechanisms of PDAC
targets, which were implicated in downstream cellular and identified key biomarkers as drug targets. This approach
dysfunctions associated with PDAC. Using the DNN-based ultimately aided in the discovery of multi-molecular drugs
DTI model trained with data from the DTI databases, the for the treatment of PDAC.
systematic drug repurposing and design process for PDAC
therapy was carried out, as shown in Figure 3. The DTI 3.3. Core signaling pathways in healthy controls
databases served as the training set for the DNN model, In healthy controls, the core signaling pathways primarily
which predicted candidate molecular drugs for PDAC drug include PI3K-AKT, TP53, and interleukin 17 (IL-17)
targets. These candidate drugs were screened according to signaling pathways. The TP53 pathway plays a crucial role
drug design specifications, including regulatory ability, in regulating the cell cycle by modulating the expression
sensitivity, and toxicity, to identify a suitable combination and activity of genes such as p21, CDK4/6, CDK2, BAX,
of molecular drugs for PDAC treatment. E2F, and MUVB. These interactions influence cell

Volume 4 Issue 1 (2025) 60 doi: 10.36922/td.4709