biostudies-literatureLau BTAmerican Cancer SocietyClayville FoundationHoward Hughes Medical InstituteDoris Duke Charitable FoundationNHGRI NIH HHSNational Cancer InstituteNCI NIH HHSSeiler FoundationNational Human Genome Research Institute11913-11917https://www.ebi.ac.uk/biostudies/studies/S-EPMC8045410biostudies-literatureUnknown89(22)Digital PCR (dPCR) relies on the analysis of individual partitions to accurately quantify nucleic acid species. The most widely used analysis method requires manual clustering through individual visual inspection. Some automated analysis methods have emerged but do not robustly account for multiplexed targets, low target concentration, and assay noise. In this study, we describe an open source analysis software called Calico that uses "data gridding" to increase the sensitivity of clustering toward small clusters. Our workflow also generates quality score metrics in order to gauge and filter individual assay partitions by how well they were classified. We applied our analysis algorithm to multiplexed droplet-based digital PCR data sets in both EvaGreen and probes-based schemes, and targeted the oncogenic BRAF V600E and KRAS G12D mutations. We demonstrate an automated clustering sensitivity of down to 0.1% mutant fraction and filtering of artifactual assay partitions from low quality DNA samples. Overall, we demonstrate a vastly improved approach to analyzing ddPCR data that can be applied to clinical use, where automation and reproducibility are critical.Analytical chemistryRobust Multiplexed Clustering and Denoising of Digital PCR Assays by Data Gridding.PMC8045410R33CA174575P01 HG000205R01 HG006137R01HG006137R33 CA174575RSG-13-297-01-TBGP01HG000205Wood-Bouwens CJi HPLau BTfalseRobust Multiplexed Clustering and Denoising of Digital PCR Assays by Data Gridding.Digital PCR (dPCR) relies on the analysis of individual partitions to accurately quantify nucleic acid species. The most widely used analysis method requires manual clustering through individual visual inspection. Some automated analysis methods have emerged but do not robustly account for multiplexed targets, low target concentration, and assay noise. In this study, we describe an open source analysis software called Calico that uses "data gridding" to increase the sensitivity of clustering toward small clusters. Our workflow also generates quality score metrics in order to gauge and filter individual assay partitions by how well they were classified. We applied our analysis algorithm to multiplexed droplet-based digital PCR data sets in both EvaGreen and probes-based schemes, and targeted the oncogenic BRAF V600E and KRAS G12D mutations. We demonstrate an automated clustering sensitivity of down to 0.1% mutant fraction and filtering of artifactual assay partitions from low quality DNA samples. Overall, we demonstrate a vastly improved approach to analyzing ddPCR data that can be applied to clinical use, where automation and reproducibility are critical.2017-01-01T00:00:00Z2017 Nov2024-11-12T15:01:01.813Z2022-02-09T14:55:00.131ZS-EPMC80454102908314310.1021/acs.analchem.7b02688