Puncture biopsy, especially those preserved by formalin fixed paraffin embedding (FFPE) samples, play an important role in various research purposes. Diverse single-nucleus RNA sequencing (snRNA-seq) techniques have been developed for FFPE samples, however, how to perform high-throughput snRNA-seq on small FFPE puncture samples is still a challenge. Here, the previously developed snRNA-seq technique (snRandom-seq) is optimized by implementing a pre-indexing procedure for the minimal puncture FFPE samples. In analyzing 20 samples from various solid tumors, optimized snRandom-seq still detected ≈17 000 genes and 12 000 long non-coding RNAs (lncRNAs), achieving precise clustering based on tissue origin. A head-to-head comparison with 10× Genomics on fresh biopsy samples showed a similar gene detection rate, with significantly enhanced lncRNA detection, indicating that the optimized snRandom-seq technique maintains its established gene detection advantages even when applied to small samples. Utilizing 7 puncture FFPE samples of liver metastases from 3 colorectal cancer patients pre- and post-immunotherapy, the cellular developmental trajectories are reconstructed and revealed dynamic spatiotemporal heterogeneity during treatment, including insights into pseudoprogression of immunotherapy. Therefore, the optimized snRandom-seq offers a solution for high-throughput single-cell RNA and non-coding RNA analysis in minimal puncture FFPE sample.