Fuzzy K-means-based outlier detection in plastic-degradation-related protein sequences using PSI-BLAST, Jaccard...

Plastic pollution is a severe environmental hazard due to the persistence of synthetic polymers, thus requiring the development of sophisticated techniques. This study proposes a fuzzy k-means-based framework integrating PSI-BLAST alignment features, Jaccard motif similarity, and OMA evolutionary scores to identify functional outliers within protein sequences associated with plastic degradation. Three variations of the system were evaluated: (i) PSI-BLAST alone could identify 200 outliers, (ii) PSI-BLAST + Jaccard similarity reduced the number of outliers to 162, and (iii) PSI-BLAST + Jaccard + OMA further reduced the number of outliers to 151, achieving a 24.5% improvement in outlier detection. The approach shows that proteins with weak similarity to known degraders, suggesting candidates for novel catalytic functions. A knowledge graph constructed from the clustering results visualises connectivity patterns and isolates weakly linked outliers. These results highlight that combining evolutionary, structural, and functional metrics improves the precision of detecting non-canonical sequences relevant to plastic degradation pathways.