Machine learning strategies to characterize and discover antimicrobial peptides
Lazar, Kathryn M.
2022
-
Antimicrobial peptide (AMP) based therapeutics are promising alternatives to broadly neutralizing antibiotics that cause antimicrobial resistance and threaten public health; however, applications involving AMPs are limited by a lack of understanding of their characteristics, particularly target specificity. Here, two predictive models were developed using a novel workflow that generates peptide ... read moredescriptors based on physiochemical properties, performs feature reduction using LASSO regression, and builds a model using a random forest machine learning algorithm. Both models demonstrate high predictive power; the first model classifies peptides as either AMPs or nonAMPs, and the second classifies AMPs based on whether they target E. coli or B. subtilis. Based on these models, 12 unique 20-mer peptides with predicted target specific antimicrobial activity were identified. Four of these peptides, along with eight controls, were evaluated via MIC assays. Three of the novel peptides demonstrated antimicrobial activity, one of which had target specificity in line with its computational prediction, warranting further investigation into its precise MIC and the effects of further modifications to the peptide sequence.
Thesis (B.S.C.H.E.)--Tufts University, 2022.
Submitted to the Dept. of Chemical and Biological Engineering.
Advisors: Kyongbum Lee, Shuchin Aeron, and Eric Miller.read less - ID:
- 9p290q80d
- To Cite:
- TARC Citation Guide EndNote
- Usage:
- Detailed Rights
