Multiomics-Driven Drug-Cell Interaction Network for Chemotherapy Sensitivity Prediction in Metabolically Defined Triple-Negative Breast Cancer Subtypes.

Triple-negative breast cancer (TNBC) is associated with a poor prognosis due to insufficient molecular subtyping precision and limited actionable targets. Although metabolic reprogramming underlies TNBC chemotherapy resistance, establishing metabolic subtyping systems and investigating drug sensitivity across distinct metabolic subgroups could provide novel therapeutic avenues for breast cancer management. GSVA (Gene Set Variation Analysis) analysis of metabolic pathways reveals significant differences in TNBC (Triple-Negative Breast Cancer) patients. TNBC patients are classified into four metabolic subtypes through consensus clustering, based on their GSVA values of metabolic pathways. These subtypes are: MS_1, characterised by increased lipogenic activity; MS_2, characterised by increased carbohydrate and nucleotide metabolism; MS_3, a metabolism-active subtype with activation of all types of metabolism; and MS_4, characterised by suppressed metabolic activity across all types of metabolism. We next propose a novel method called MODIN (Multiomics-Driven Drug-Cell Interaction Network), which embeds multi-omics gene information (mRNA expression, copy number variation and DNA methylation) and drug SMILES data into a latent space, and then employs a multi-head attention-based interaction module to accurately predict the LN_IC50 values of 621 drugs in TNBC. Based on MODIN, noteworthy disparities in drug sensitivity emerge between the patient cohorts categorised as MS_2 and MS_3. MS_3 patients show a significantly higher sensitivity to chemotherapy regimens, especially for doxorubicin and docetaxel, while the MS_2 cohort displays marked resistance to these drugs. Our study reveals the metabolic heterogeneity of TNBC, and TNBC patients with increased carbohydrate and nucleotide metabolism exhibit the poorest prognoses and greater resistance to doxorubicin and docetaxel.