Background Treating patients with advanced sarcomas is challenging due to great histologic diversity among different subtypes. Leiomyosarcoma (LMS) and liposarcoma (LPS) are the most common histologic types of soft tissue sarcoma (STS). Although both LMS and de-differentiated LPS (DDLPS) are relatively aggressive and have a high rate of recurrence, they have significant differences in histology, biology, and clinical behaviors. Methods 32 STS (20 LMS and 12 DDLPS) samples and matched peripheral blood were collected from Zhongshan Hospital, Fudan University with patient consents. WES was conducted on all the tumor/blood pairs. Somatic single nucleotide variations (SNV) and indel were identified using Genome Analysis Toolkits (GATK). Recurrent somatic copy-number alterations (SCNA) were called using GISTIC2.0. RNA-Seq was further conducted, then their chromosome rearrangement was revealed using STAR-Fusion. Next, differential gene expression analysis and pathway enrichment analysis were performed using DeSeq2 and NetwokAnalyst. Stromal and immunological landscape profiling was conducted using MCP-counter. Results WES identified different genetic variation patterns in LMS and DDLPS. Amplification of oncogenes is prominent in DDLPS, while loss of function mutation and deletion of tumor suppressors mostly occur in LMS. A focal amplification affecting chromosome 12q13–15 region which encodes MDM2, CDK4 and HMGA2, is notable in DDLPS. Mutations in tumor suppressors such as TP53, ATRX, PTEN, and RB1 are identified in LMS. RNA-seq confirmed overexpression of MDM2, CDK4 and HMGA2 in DDLPS and down-regulation of TP53 and RB1 in LMS. It also revealed different fusion patterns between LMS and DDLPS. The rate of gene fusion is significantly higher in DDLPS than in LMS (p=0.007) and fusion transcripts involving chromosome 12 can only be found in DDLPS. Stromal and immunological profiling indicate that microenvironmental signatures are highly correlated with histologic types, and DDLPS have more endothelial cell and fibroblast content than LMS. Conclusions Our analysis revealed recurrent genetic variation in LMS and DDLPS. DDLPS harbor frequent fusion events and may produce neoepitopes, which can be targeted to produce safer CAR-T (Chimeric Antigen Receptor T) cells for immunotherapy. Additionally, our study suggested the tumor microenvironment of sarcoma can potentially be correlated to patients' clinical outcomes and used to guide personalized treatment.