Recently, we had reported a synthetic positively charged leucine-rich 14-residue-long antimicrobial peptide (AMP, LL-14: NH-LKWLKKLLKWLKKL-CONH), which was highly active and cytotoxic relative to its valine analogue (VV-14). However, the thermodynamics underlying this differential toxicity and antimicrobial activity was unclear. Understanding the energetics of peptide binding to micelles (simplest membrane mimic, viz., SDS as a bacterial membrane and DPC as a eukaryotic membrane) and the effect of Leu → Val peptide mutations on the stability of the peptide:micelle complexes are of great academic interest and relevant for the rational design of potent and selective AMPs for therapeutic use. Here, we have reported the molecular dynamics free energy simulations that allowed us to quantitatively estimate the strength of peptide discrimination (based on single- or multiple-site Leu/Val mutations in LL-14) by membrane mimetic micelles (SDS and DPC) and decipher the energetics underlying peptide selectivity by micelles. The Leu-containing peptide (LL-14) was found to be preferred for micelle (SDS and DPC) binding relative to its Val analogues (single or multiple Val mutants). The strength of the preference depended on the position of the Leu/Val mutation in the peptide. Surprisingly, the N-terminal LL-14 single mutation (Leu → Val: L1V) was found to fine-tune the electrostatic interactions, resulting in the highest peptide selectivity (ΔΔ ∼ 8 kcal/mol for both SDS and DPC). However, the mechanism of L1V peptide selectivity was distinctly different for SDS and DPC micelles. SDS ensured high selectivity by disrupting the peptide:micelle salt bridge, whereas DPC desolvated the broken-peptide-backbone hydrogen bond in the V1 peptide:micelle complex. Mutations (Leu → Val) in the middle positions of the LL-14 (4th, 7th, 8th, and 11th) were disfavored by the micelles primarily due to the loss of peptide:micelle hydrophobic interactions. Peptides differing at the C-terminal (i.e., L14V) were recognized by SDS micelles (ΔΔ ∼ 4 kcal/mol) by altering peptide:micelle interactions. L14V mutation, on the other hand, did not play any role in the peptide:DPC binding, as no direct interactions between the C-terminal and DPC micelle were observed due to obvious electrostatic reasons. The strength of selectivity favoring LL-14 binding against VV-14 was found to be much higher for DPC micelles (ΔΔ ∼ 25 kcal/mol) relative to SDS micelles (ΔΔ ∼ 19 kcal/mol). The loss of the peptide:micelle hydrophobic contact in response to LL-14 → VV-14 mutation was found to be significantly larger for DPC relative to SDS micelles, resulting in higher discriminatory power for the former. Peptide:SDS salt bridges seemed to prevent the loss of peptide:micelle hydrophobic contact to some extent, leading to weaker selectivity for SDS micelles. High selectivity of DPC micelles provided an efficient mechanism for VV-14 dissociation from DPC micelles, whereas low-selectivity of SDS micelles ensured binding of both LL-14 and VV-14. To the best of our knowledge, this is the first study in which the experimental observations (antimicrobial activity and toxicity) between leucine-rich and valine-rich peptides have been explained by establishing a direct link between the energetics and structures.