If by duplicates you mean two biological replicates per condition, then yes, DESeq2 can handle that, and many studies do analyze data with n = 2 per condition. However, three or more biological replicates are generally recommended. With only two replicates, variance estimation is less stable, so statistical power is usually lower. In practice, that makes it harder to detect smaller effect sizes and to distinguish true biological signal from outliers or unexpected sample-to-sample variability. So duplicates are often usable, but (if feasible) triplicates (or more) are preferred. If by duplicates you mean technical replicates, that is a different issue, since technical replication is not a substitute for biological replication.

Also, if additional samples are collected across different batches, any batch effects should be accounted for in the design formula when possible.