I was working in subtractive proteomics to discover potential vaccine candidates from the entire bacterial proteome to construct effective vaccines. So, during my study and research, I came across the broader term "homologous proteins". It is very interesting to identify and cluster homologs across the proteome, then select one as a representative sequence. There are many methods and tools, such as CD-HIT analysis, for this task. But can we classify homologs (paralogs and orthologs) only based on sequence similarity, or by sequence identity, using any alignment tool like BLAST or Diamond? If so, what threshold values should be preferred?