MuTect is a method for detecting somatic point mutations in impure and heterogeneous cancer samples. It uses a Bayesian classifier to identify mutations with very low allele fractions, requiring only a few supporting reads, followed by carefully tuned filters to ensure high specificity. The method is benchmarked using real sequencing data to evaluate sensitivity and specificity as a function of sequencing depth, base quality, and allelic fraction. Compared to other methods, MuTect has higher sensitivity with similar specificity, especially for mutations with allelic fractions as low as 0.1 and below, making it particularly useful for studying cancer subclones and their evolution in standard exome and genome sequencing data. Somatic single-nucleotide substitutions are important and common mechanisms for altering gene function in cancer. However, they are difficult to identify due to their low frequency in the genome and the presence of mutations in only a small fraction of DNA molecules. The study of subclonal structure is critical for understanding tumor evolution and developing clinical diagnostic tools for personalized cancer therapy. Recent studies on subclonal events in cancer have used nonstandard experimental strategies, including analysis of clonal mutations in metastases, ultra-deep sequencing, and sequencing of single cells. In contrast, tens of thousands of tumors are sequenced at standard depths as part of large-scale cancer genome projects. To detect clonal and subclonal mutations in these samples, a highly sensitive and specific mutation-calling method is needed. The sensitivity and specificity of somatic mutation-calling methods vary depending on factors such as sequence coverage, base quality, allelic fraction, and evidence thresholds. Characterizing these factors is necessary for designing experiments with adequate power to detect mutations at a given allelic fraction. To meet these needs, MuTect was developed. It uses a Bayesian classifier to detect mutations with very low allele fractions and applies filters to ensure high specificity. MuTect is publicly available and has been validated in multiple studies, showing high sensitivity and specificity for detecting low-allelic-fraction events. MuTect is freely available for noncommercial use at http://www.broadinstitute.org/cancer/cga/mutect. It is particularly useful for analyzing samples with low purity or complex subclonal structures. The method has been benchmarked using real sequencing data and has shown higher sensitivity than other methods for low-allelic-fraction events while maintaining high specificity. This makes MuTect a valuable tool for studying cancer subclones and their evolution in standard exome and genome sequencing data.MuTect is a method for detecting somatic point mutations in impure and heterogeneous cancer samples. It uses a Bayesian classifier to identify mutations with very low allele fractions, requiring only a few supporting reads, followed by carefully tuned filters to ensure high specificity. The method is benchmarked using real sequencing data to evaluate sensitivity and specificity as a function of sequencing depth, base quality, and allelic fraction. Compared to other methods, MuTect has higher sensitivity with similar specificity, especially for mutations with allelic fractions as low as 0.1 and below, making it particularly useful for studying cancer subclones and their evolution in standard exome and genome sequencing data. Somatic single-nucleotide substitutions are important and common mechanisms for altering gene function in cancer. However, they are difficult to identify due to their low frequency in the genome and the presence of mutations in only a small fraction of DNA molecules. The study of subclonal structure is critical for understanding tumor evolution and developing clinical diagnostic tools for personalized cancer therapy. Recent studies on subclonal events in cancer have used nonstandard experimental strategies, including analysis of clonal mutations in metastases, ultra-deep sequencing, and sequencing of single cells. In contrast, tens of thousands of tumors are sequenced at standard depths as part of large-scale cancer genome projects. To detect clonal and subclonal mutations in these samples, a highly sensitive and specific mutation-calling method is needed. The sensitivity and specificity of somatic mutation-calling methods vary depending on factors such as sequence coverage, base quality, allelic fraction, and evidence thresholds. Characterizing these factors is necessary for designing experiments with adequate power to detect mutations at a given allelic fraction. To meet these needs, MuTect was developed. It uses a Bayesian classifier to detect mutations with very low allele fractions and applies filters to ensure high specificity. MuTect is publicly available and has been validated in multiple studies, showing high sensitivity and specificity for detecting low-allelic-fraction events. MuTect is freely available for noncommercial use at http://www.broadinstitute.org/cancer/cga/mutect. It is particularly useful for analyzing samples with low purity or complex subclonal structures. The method has been benchmarked using real sequencing data and has shown higher sensitivity than other methods for low-allelic-fraction events while maintaining high specificity. This makes MuTect a valuable tool for studying cancer subclones and their evolution in standard exome and genome sequencing data.