iUmami-SCM

iUmami-SCM predicts umami-active peptides from primary sequence information by applying a scoring card method to identify sequence determinants of umami taste.

Key Features:

• Sequence-Based Prediction: Uses primary sequence information with a scoring card method (SCM) that incorporates propensity scores of amino acids and dipeptides to identify potential umami peptides.
• Predictive Performance: Reports an accuracy of 0.865 and a Matthews correlation coefficient of 0.679 on independent datasets and exhibits superior performance compared to other machine learning classifiers.
• Biophysical and Biochemical Insight: Provides SCM-derived propensity scores that give insights into biophysical and biochemical properties associated with umami taste intensity.
• Benchmark Datasets: Evaluated using datasets UMP-TR (112 umami and 241 non-umami peptides) and UMP-IND (28 umami and 61 non-umami peptides).

Scientific Applications:

• Umami peptide identification: Identifies candidate umami-active peptides from sequence data for downstream validation.
• Flavor enhancement and product development: Supports selection of peptide candidates for flavor enhancement and food product formulation.
• Computational screening alternative: Serves as a computational alternative to experimental assays for high-throughput screening, reducing time and cost in umami peptide discovery.

Methodology:

Applies a scoring card method (SCM) that computes propensity scores for amino acids and dipeptides to evaluate sequences; the approach was evaluated on UMP-TR (112 umami, 241 non-umami) and UMP-IND (28 umami, 61 non-umami) and compared to other machine learning classifiers, reporting accuracy 0.865 and MCC 0.679 on independent data.