Locsmoc is a tool for generating piecewise polynomial models for one-dimensional signals. It is designed to detect and characterize local trends in data, trend changes, and discontinuities. It is suitable for producing polynomial representations of genomic tracks such as microarray copy-number data, coverage tracks from high-throughput sequencing, or any other signals associating a scalar value with genomic coordinates.
Locsmoc reads, processes, and outputs signals in run-length encoded form. This makes it conceptually and technically different from many other model-producing or smoothing algorithms such as regression, splines, loess, etc. that are based on point-wise data. The core algorithm works stochastically from the bottom up, considering pairs of neighboring input data and joining them into longer segments to produce a large scale model.
The input and output format are described in detail here. Some usage examples are shown here. The tool is written in Java and uses some third-party libraries. These libraries are included in the default download file.
M. Tarabichi, V. Detours, and T. Konopka. Piecewise polynomial representations of genomic tracks. PLOS ONE 7(11) e48941, 2012.