Gapfilling

Gap filling is the process of inferring and adding missing data elements in a dataset, sequence, or model to produce a more complete or usable representation. Gaps can arise from incomplete data collection, sequencing errors, or intentional redaction. Different domains tailor gap filling to their data types and objectives, balancing accuracy with parsimony.

In text processing and natural language processing, gap filling can refer to restoring missing words or characters in corrupted documents, or generating plausible substitutes in cloze tasks. Methods range from rule-based heuristics to statistical imputation, language models, and context-aware prediction.

In genomics and bioinformatics, gap filling commonly refers to closing gaps in a genome assembly. After initial assembly, gaps between contigs or scaffolds may be closed using sequencing reads that span the gaps, paired-end information, or targeted re-sequencing. Tools attempt to fill the gaps with consensus bases or insert long contigs, improving continuity and assembly metrics; some gaps may remain if data are insufficient.

In constraint-based metabolic network reconstructions, gap filling identifies missing reactions that prevent a model from achieving growth or other objectives. The process adds reactions from a curated universal set, typically subject to optimization and biological plausibility constraints, to restore network functionality while keeping the added content minimal.

Gap filling also appears in time series, image, and sensor data, where missing observations are imputed using interpolation, regression, or machine learning to produce continuous sequences for analysis.