1. The summary diagram of the study

2. Main analysis steps in the study

2.1 Propose computational algorithm based on Complementary signature regulation

• 1）TCGA-DEG identification and Over-representation enrichment analysis;
• 2）RAR proximity of Signature to breast cancer targets;
• 3）De-redundancy analysis between signatures;
• 4）Propose wAC index based on reversal regulation to recognize marker signature.

2.2 Propose computational algorithm based on Complementary signature regulation

• 1）data collection and cleaning；
• 2）AutoML model training；
• 3）Model comparison and selection.

2.3 Discover TCM compound combination

• 1）Signature regulation score of single TCM compound；
• 2）Signature regulation score of TCM compound combinations；
• 3）Overall ML prediction based on ZIP synergistic measurements；
• 4）Validation of synergy effect via a series of in vitro experiments

3. Public datasets used in the study

• Gene signatures: MSigDB, https://www.gsea-msigdb.org/gsea/msigdb/index.jsp
• Drug--Disease, Drug--Target, Target--Disease information: https://db.idrblab.net/ttd/full-data-download
• Protein–Protein Interactions: https://www.nature.com/articles/s41467-019-09186-x#Sec23, Supplementary Data 1
• Compound-perturbating signatures ;Metadata for signatures, genes, compounds: Expanded CMap LINCS Resource 2020 , https://clue.io/data/CMap2020#LINCS2020
• GDSC1-dataset, GDSC2-dataset: GDSC, https://www.cancerrxgene.org/downloads/bulk_download
• TCM High-throughput Data, TCM metadata: ITCM, http://itcm.biotcm.net/download.html
• Drug Combination Measurement: SYNERGxDB, https://www.synergxdb.ca/synergy_score?&dataset=2
• Breast cell lines mRNA expression: CCLE, https://depmap.org/portal/download/all/