MapSplice is a second-generation algorithm for accurately mapping RNA-seq reads to splice junctions, focusing on high sensitivity and specificity. It works with both short (<75 bp) and long reads (≥75 bp) and does not rely on splice site features or intron length, enabling detection of both canonical and non-canonical splices. It leverages the quality and diversity of read alignments to improve accuracy. MapSplice outperforms TopHat and SpliceMap in sensitivity and specificity on simulated RNA-seq data. It was validated using eight breast cancer RNA-seq datasets, showing global consistency in alternative splicing and differences between molecular subtypes. MapSplice is efficient in CPU and memory usage and can handle both short and long reads. It was used to analyze alternative splicing differences between basal and luminal breast cancer subtypes, with qRT-PCR validation showing high correlation between isoform proportions and splice junction estimates. MapSplice detected a wide range of splice junctions, including many not previously observed in full-length transcripts. It demonstrated high accuracy in detecting splice junctions, with over 98% specificity and 96% sensitivity in simulated data. MapSplice's performance was validated using synthetic data and experimental results from breast cancer samples. It showed high concordance with previous studies and accurately identified alternative splicing events. The algorithm is efficient, scalable, and suitable for both model organisms and organisms with sparse transcript annotations. It is a powerful tool for analyzing alternative splicing in RNA-seq data.MapSplice is a second-generation algorithm for accurately mapping RNA-seq reads to splice junctions, focusing on high sensitivity and specificity. It works with both short (<75 bp) and long reads (≥75 bp) and does not rely on splice site features or intron length, enabling detection of both canonical and non-canonical splices. It leverages the quality and diversity of read alignments to improve accuracy. MapSplice outperforms TopHat and SpliceMap in sensitivity and specificity on simulated RNA-seq data. It was validated using eight breast cancer RNA-seq datasets, showing global consistency in alternative splicing and differences between molecular subtypes. MapSplice is efficient in CPU and memory usage and can handle both short and long reads. It was used to analyze alternative splicing differences between basal and luminal breast cancer subtypes, with qRT-PCR validation showing high correlation between isoform proportions and splice junction estimates. MapSplice detected a wide range of splice junctions, including many not previously observed in full-length transcripts. It demonstrated high accuracy in detecting splice junctions, with over 98% specificity and 96% sensitivity in simulated data. MapSplice's performance was validated using synthetic data and experimental results from breast cancer samples. It showed high concordance with previous studies and accurately identified alternative splicing events. The algorithm is efficient, scalable, and suitable for both model organisms and organisms with sparse transcript annotations. It is a powerful tool for analyzing alternative splicing in RNA-seq data.