MORF: A Computational Framework for Synonymous Optimisation of Co-Encoding Regions in Protein-Coding Sequences

The limited packaging capacity of adeno-associated virus (AAV) vectors presents a fundamental constraint in gene delivery, restricting the amount of genetic information that can be encoded within a single construct. Motivated by this limitation, this work investigates a computational framework for co-encoding multiple protein-coding sequences within a single nucleotide sequence through the use of overlapping reading frames (RFs). This paper formally defines the co-encoding problem and proposes an algorithmic solution in the form of the multiple overlapping reading frames (MORF) tool. The approach first employs a sliding Hamming similarity (SHS) function to identify candidate alignments between two protein-coding nucleotide sequences. Rather than retaining only the highest-scoring alignment, a set of leading candidate co-encoding regions (CERs) is selected for subsequent optimisation, allowing configurations with lower initial similarity that may achieve greater nucleotide-level similarity following optimisation to be explored. Within each candidate CER, synonymous codon substitutions provide a finite set of admissible alternatives for each encoded amino acid, enabling the nucleotide sequences to be modified while preserving the corresponding protein sequences. Each candidate is optimised using a sequential two-pass procedure. The query sequence is first modified through synonymous substitution of its native codons while the reference sequence is held fixed, with candidate codons evaluated according to their agreement with corresponding aligned three-nucleotide windows. The resulting modified query sequence is subsequently held fixed while synonymous substitutions are evaluated for the reference sequence in the same manner. Only substitutions producing a strict improvement in local nucleotide-level agreement are retained. Following optimisation of all retained candidates, the resulting configurations are compared and the candidate achieving the greatest post-optimisation nucleotide-level similarity is selected. The framework is evaluated using the coding sequences of Green Fluorescent Protein (GFP) and DsRed as a proof-of-concept pair. Optimisation of the highest-ranked initial SHS configuration increases nucleotide-level similarity within the CER from 32.96% to 51.41%. Extending the optimisation across the retained candidate configurations identifies a lower-ranked initial configuration whose similarity increases from 32.17% to 60.05%, demonstrating that the initially highest-scoring alignment does not necessarily achieve the greatest similarity following optimisation. These results establish the computational feasibility of increasing nucleotide-level agreement between overlapping protein-coding sequences through synonymous codon substitution while preserving their encoded amino-acid sequences, providing a foundation for the development of compact multi-protein genetic constructs.