Data-driven de novo design of super-adhesive hydrogels

Hydrogel fabrication

All copolymer gels have been synthesized by one-step free-radical copolymerization of monomers with a chemical crosslinker. The crosslinker focus was fastened at 0.1 mol% relative to the overall monomer content material to stability the elasticity and deformability of the gels²⁷. DMSO options containing useful monomers (complete focus of two.4 M) with compositions derived from DM and ML (Supplementary Tables 2 and seven), chemical crosslinker (glycerol 1,3-diglycerolate diacrylate, 2.4 mM), and UV initiator (2-oxoglutaric acid, 6 mM) have been used. For instance, to arrange the G-max gel, 1.819 g of BA, 0.413 g of HEA, 0.264 g of CBEA, 0.561 g of ATAC, 0.441 g of PEA, 8.4 mg of glycerol 1,3-diglycerolate diacrylate and eight.8 mg of 2-oxoglutaric acid have been added to a ten ml volumetric flask, adopted by DMSO to achieve 10 ml. The precursor resolution was transferred to a glove field to take away oxygen, poured right into a response cell (two 10 cm × 10 cm glass plates, 0.5-mm spacing) and irradiated with UV gentle (365 nm wavelength, 4 mW cm⁻² depth) for 8 h to type gels (Supplementary Fig. 9a). After UV irradiation, over 99% of the monomers have been transformed into polymers, as confirmed by NMR (Supplementary Fig. 9b).

The as-prepared organogels have been then immersed in regular saline (0.154 M NaCl) to take away solvent and residual chemical substances, with the saline exchanged each 12 h for a minimum of 2 weeks till swelling equilibrium was reached. Hydrogels have been saved in regular saline earlier than use.

Underwater adhesion characterization

The tack check was performed utilizing a SHIMADZU tester (Autograph AG-X) outfitted with Trapezium X software program. Hydrogel (0.3–0.8 mm thickness) at swelling equilibrium was adhered to the probe utilizing cyanoacrylate adhesive (tremendous glue). For fast screening, DM-driven hydrogels from the coaching spherical and ML-driven hydrogels from three optimization rounds, have been ready as 15 mm diameter samples. For detailed adhesion research, 10 mm diameter samples have been used to keep away from exceeding the drive vary of the instrument. This change in diameter didn’t have an effect on the adhesive energy outcomes. The hydrogel on the probe was then immersed in a check resolution (for instance, regular saline) for five min to achieve equilibrium. The probe descended in direction of the substrate at 1 mm min⁻¹ till a loading drive of 10 N was utilized, maintained for 10 s and withdrawn at 10 mm min⁻¹ (Supplementary Fig. 10). These check situations have been used as a regular protocol until in any other case specified. For repeated adhesion checks, hydrogels rested underwater for five min between cycles, with glass substrates changed each 100 checks. For extended attachment–detachment cycles (Extended Data Fig. 8), a 5 N loading drive and a ten s contact time have been used to attenuate gel fatigue. Each pattern was examined a minimum of 3 times. For hydrogel dataset building, the very best adhesive energy recorded for every pattern was reported as F_a, representing most adhesion efficiency below the particular situations.

Lap shear adhesive energy was measured utilizing a common testing machine (UTM, INSTRON 5965). A hydrogel (10 mm diameter, space A = 78.5 mm²) at swelling equilibrium was sandwiched between two glass slides, pressed at 20 N for 1 min in regular saline. Shear loading was utilized at 50 mm min⁻¹. Shear adhesive energy (F_a) was calculated as F_a = F_max/A, the place F_max is the utmost loading drive. For adhesion sturdiness checks (Supplementary Fig. 15), the sandwiched meeting was saved in regular saline for various durations earlier than testing.

Interfacial toughness was measured by 180° peeling checks utilizing INSTRON 5965. Hydrogel strips (10 mm × 150 mm) have been adhered to a glass substrate in regular saline utilizing gentle finger strain, adopted by a 2 kg hand curler utilized in every course for 1 min to make sure uniform contact. Polyethylene terephthalate (PET) movies (50 μm thickness) served as a stiff backing. Peeling checks have been performed at 50 mm min⁻¹. Interfacial toughness (G_c) was calculated as G_c = 2F_c/w, the place F_c is the plateau drive and w is the pattern width (10 mm).

DM of adhesive proteins

A complete dataset of adhesive proteins was compiled from the NCBI protein database, utilizing ‘adhesive proteins’ because the question key phrase. A complete of 24,707 protein sequences from 3,822 totally different organisms (micro organism, viruses, eukaryotes and animals) have been collected with out further knowledge cleansing. Based on taxonomy annotations, proteins have been grouped by species, and a consensus sequence was generated for every species to seize widespread sequence patterns and scale back the affect of particular person variations.

The dataset included 3,111 species, noting that taxonomic overlap ends in protein counts not summing to 24,707. For strong evaluation, the highest 200 species, ranked by the variety of distinct proteins recognized per species, have been chosen for additional research.

Protein sequences have been exported in FASTA format⁴⁵ utilizing the Bio.SeqIO interface in BioPython⁴⁶. Consensus sequences have been computed with Clustal Omega²³, which performs a number of sequence alignment by producing a distance matrix from pairwise alignments, setting up a information tree primarily based on evolutionary relationships and progressively aligning sequences from the closest to probably the most distant. The ensuing alignment identifies probably the most frequent residues at every place, yielding a consensus sequence that highlights conserved areas.

Clustal Omega was executed with the command:

the place “input_file” and “output_aln_file” denote the enter protein sequences and output consensus sequences, respectively. The 200 consensus sequences generated have been used for subsequent sequence evaluation and hydrogel formulation design.

ML strategies

A six-dimensional function vector, ϕ_i = [ϕ_BA, ϕ_HEA, ϕ_CBEA, ϕ_ATAC, ϕ_AAm, ϕ_PEA], was used to symbolize monomer proportions in hydrogels. The goal variable was adhesive energy, F_a. To mannequin the connection between ϕ_i and F_a, we explored each linear and non-linear ML fashions (Supplementary Tables 5 and 6).

Linear fashions included least absolute shrinkage and choice operator regression (Lasso) and ridge regression (Ridge). Non-linear fashions comprised ok-nearest neighbours (KNN), kernel ridge regression (KRR), assist vector regression (SVR), random forest regression (RFR), gradient boosting regression with XGBoost (XGB), additional bushes regression (ETR) and Gaussian course of (GP) with a Matérn kernel^32,34. These non-linear fashions embody non-parametric (KNN), kernel-based (KRR, SVR and GP) and tree-ensemble (RFR, XGB and ETR) approaches, enabling a complete comparability^34,35,47.

XGB was of v.1.6.2, whereas the opposite fashions have been applied utilizing Scikit-learn (v.1.0.2) and Scikit-optimize (v.0.9.0). The hyperparameter n_estimators was tuned utilizing Optuna⁴⁸, whereas others have been optimized utilizing grid search (Supplementary Table 6). A ten-fold cross-validation technique was used to evaluate predictive efficiency on our dataset of 180 hydrogels, utilizing root imply squared error (RMSE) because the metric. GP and RFR, with the bottom RMSE in training-test error utilizing a 90%/10% practice/check cut up (Extended Data Fig. 4), emerged as the highest performer and runner-up, respectively, and have been subsequently used as the bottom (surrogate) fashions.

To make extrapolative predictions, we tried three kinds of strategies.

1. 1.

   Exploitation-only enumeration:

   Ten million ϕ_i vectors have been generated from a uniform distribution [0, 1.0) for each monomer, normalized to sum to 1.0. The top five vectors, ranked by predicted F_a from each model, were experimentally validated.

2. 2.

   Batched BO:

   • GP_KB: used GP predictions as the hypothetical values for selecting the next data points maximizing EI.

   • GP_CLmax: used the maximum F_a (y_max) from the training set as a hypothetical value for selecting the next data points with EI maximums.

   • GP_CLmin: used the minimum F_a (y_min) for selecting the next data points with EI maximums.

   • GP_LP: incorporated a locally penalized term in EI calculation³⁷.

   GP_KB, GP_CLmax and GP_CLmin simplified the joint q-EI probability calculation³⁶ by using the GP prediction value as a hypothetical value for selecting the next data points with EI maximums. A batch size of q = 10 was selected.

3. 3.

   Batched sequential model-based optimization (SMBO):

   • GP-RFR: GP as the hypothetical value provider and RFR as the EI maximizer.

   • RFR-RFR: RFR as both the hypothetical value provider and the EI maximizer.

   • RFR-GP: RFR as the hypothetical value provider and GP as the EI maximizer.

   • RFR-GP*: RFR-GP with a warm start, 10 RFR-generated points were added to the real dataset for GP regression.

   • RFR-ETR: RFR as the hypothetical value provider and ETR as the EI maximizer.

   • RFR-GBM: RFR as the hypothetical value provider and GBM as the EI maximizer.

   SMBO iteratively updates the surrogate model while exploring promising data points³³. GP and RFR, when used as the hypothetical value providers, balance exploitation and exploration, whereas GP_CLmax and GP_CLmin emphasize exploitation and exploration, respectively⁴⁹.

SMBO (Supplementary Algorithm 1) consists of four components: the true function (f), global domain (X), acquisition function (S) and surrogate model (M). Initial training data (D) are sampled from X, and experimental F_a values are obtained (line 1). The surrogate model M is fitted to D (line 3) and S (EI) identifies the next data point based on predictive uncertainty (line 4). This data point is subsequently validated experimentally (line 5), updating D (line 6) for T iterations (line 2).

EI quantifies expected improvement, ({int }_{y* }^{infty }(y-{y}^{* })p(y){rm{d}}y), over the current best target (y^*). Owing to the time-intensive nature of hydrogel fabrication (each takes about 2 weeks), GP and RFR were used as the hypothetical value providers, enabling the maximization of the joint q-EI probability without requiring new experiments per iteration. EI maximizers (GP, RFR, ETR and GBM) used hyperparameters from Scikit-optimize (v.0.9.0).

For GP as the EI maximizer, the limited-memory Broyden–Fletcher–Goldfarb–Shannon (L-BFGS-B) algorithm⁵⁰ was executed 20 times per iteration (40 iterations total) to identify the point with the highest EI, updating the GP prior. For the other three EI maximizers (RFR, ETR and GBM), 10,000 points were randomly sampled per iteration, as numerical optimization is more suitable for tree-ensemble models lacking gradient information. SMBO ran for 40 iterations with each EI maximizer, selecting two sets of 10 data points in each iteration: the top 10 ranked by EI values (batch size q = 10), and the top 10 ranked by predicted F_a values for experimental validation. These two sets may overlap, and the total number of data points may be less than 20.

For BO methods (GP_KB, GP_CLmax, GP_CLmin and GP_LP), the procedure was similar, except that the hypothetical value provider was either GP itself (GP_KB and GP_LP) or constant values (y_max for GP_CLmax and y_min for GP_CLmin).

After the first round, 109 validated points expanded the dataset to 289 hydrogels. The second and third rounds added 27 and 25 points, respectively, resulting in a final dataset comprising 341 hydrogels.