Dynamic Methods for Missing Value Estimation for DNA Sequences

Many gene expressions in the DNA micro array and gene sequences have missing values in the datasets. It is critical to estimate these missing values accurately, because most of the existing algorithms for gene expression analysis require the complete DNA dataset as an input, which affects the performance of gene classifications. This paper introduces dynamic local least squares imputation (DLLSimpute), which selects local matrices dynamically and consequently uses more gene information each time to recover the missing values. Numerical results show that the DLLSimpute recovers missing values more accurately than the k-nearest neighboring imputation (KNNimpute) and the local least squares imputation (LLSimpute) regardless of the completeness of the datasets.