Aggregation Rules of Short Peptides

Abstract

The elucidation of aggregation rules for short peptides (e.g., tetrapeptides and pentapeptides) is crucial for the precise manipulation of aggregation. In this study, we derive comprehensive aggregation rules for tetrapeptides and pentapeptides across the entire sequence space based on the aggregation propensity values predicted by a transformer-based deep learning model. Our analysis focuses on three quantitative aspects. First, we investigate the type and positional effects of amino acids on aggregation, considering both the first- and second-order contributions. By identifying specific amino acids and amino acid pairs that promote or attenuate aggregation, we gain insights into the underlying aggregation mechanisms. Second, we explore the transferability of aggregation propensities between tetrapeptides and pentapeptides, aiming to explore the possibility of enhancing or mitigating aggregation by concatenating or removing specific amino acids at the termini. Finally, we evaluate the aggregation morphologies of over 20,000 tetrapeptides, regarding the morphology distribution and type and positional contributions of each amino acid. This work extends the existing aggregation rules from tripeptide sequences to millions of tetrapeptide and pentapeptide sequences, offering experimentalists an explicit roadmap for fine-tuning the aggregation behavior of short peptides for diverse applications, including hydrogels, emulsions, or pharmaceuticals.

Figure 1

Analyses of the aggregation rules of short peptides from three perspectives. A deep learning-based TRN model predicts the AP values of tetrapeptides (AP_te) and pentapeptides (AP_pe) across the entire sequence space (i.e., 160,000 tetrapeptides and 3,200,000 pentapeptides). Based on the predicted AP values, we analyze the (a) type and positional contribution of 20 amino acids to the AP, (b) transferability relationship between the AP of a pentapeptide and the averaged AP of two corresponding tetrapeptides (AP_pe versus AP_ave), and (c) distribution of the morphologies of aggregated tetrapeptides as well as the type and positional contributions of 20 amino acids to different morphologies.

Figure 2

Prediction and distribution of AP_te (and AP_te′) and AP_pe (and AP_pe′). (a,b) Predicted AP values of tetrapeptides (AP_te) and pentapeptides (AP_pe) compared to simulation-generated AP of tetrapeptides (AP_sim,te) and pentapeptides (AP_sim,pe). (c,d) Violin distribution of AP_te′ (normalized AP_te) and AP_pe′ (normalized AP_pe) within four ranges of A ∈ [0.00,0.25), B ∈ [0.25,0.50), C ∈ [0.50,0.75), and D ∈ [0.75,1.00), with the number of peptides counted in each range. The purple and yellow dots overlapping the “Total” distribution in (c) indicate the comparison results with experimental TEM images, i.e., consistent and inconsistent on a qualitative level, respectively.

Figure 3

Aggregation rules in terms of individual amino acids (i.e., first-order aggregation rules). (a,b) Percentage of each amino acid at each position within the pool of aggregating peptides (i.e., the normalized AP values within the range D ∈ [0.75, 1.00]) for (a) tetrapeptides and (b) pentapeptides. (c,d) Averaged AP values of peptides with a fixed amino acid at a specific position for (c) tetrapeptides and (d) pentapeptides. (e,f) Averaged AP values of peptides containing (e) 1–4 and (f) 1–5 specific amino acids.

Figure 4

Aggregation rules in terms of 400 amino acid pairs (i.e., second-order aggregation rules). (a,b) Percentage of amino acid pairs in the high AP range of D (AP′ ∈ [0.75, 1.00]) for (a) tetrapeptides and (b) pentapeptides. (c,d) Averaged AP values of (c) tetrapeptides and (d) pentapeptides with amino acid pairs fixed at specific “positions”.

Figure 5

Transferability relationship between AP_pe′ and averaged AP_ave′ (AP_ave′ are the average of two AP_te′). (a) Relationship between AP_pe′ and AP_ave′, with the color indicating the difference ε between AP_pe′ and the averaged AP_ave′ (ε = AP_pe′ – AP_ave′). (b) Violin distribution (left) and number distribution (right) of ε. (c–e) AP values of 10 groups of peptides with positive maximum ε between AP_pe and AP_ave. (f–h) AP values of the 10 groups of peptides with negative maximum ε between AP_pe and AP_ave.

Figure 6

Computational and experimental results of AP and morphologies of peptide Aβ_16–22 with the addition of amino acids DD and DDD at the N-terminus. The table lists the attribution of each amino acid with the increasing number of D, as well as the predicted (AP_PRD) and simulation AP (AP_SIM).

Figure 7

Distribution of morphologies after aggregation. (a–c) Possible morphologies formed in simulations, such as fibers or tubes, intermediate structures such as net and curved sheet, and spherical or vesicle structures. (d) Violin distributions of morphologies. (e–g) Positional percentage of each amino acid within each RMOI range, calculated through dividing the number of each amino acid at each position within a RMOI range, by the total number of the amino acid at the position across all ranges.