Protein-based materials: Applications, modification and molecular design

Abstract

Proteins are the fundamental building blocks of nature, constructing complex molecular machines and dynamic materials. They form the protein complexes that drive life cycles and the cellular skeletal components and muscle fibers responsible for movements. Due to their extensive molecular diversity, many biomedical and industrial challenges associated with natural proteins remain unsolved. This review presents a comprehensive analysis of the structure and function of fibrous proteins, elastin, and mucins, emphasizing their roles as protein materials. It also explores their diverse applications across food, environmental, and biomedical sectors. Additionally, we focus on strategies for optimizing protein structure, including chemical modifications and molecular design, by comparing current design methods and software to summarize recent technological developments. Finally, we explore the challenges and prospects of applying artificial intelligence to complex protein structure design. This application has the potential to advance the development and application of intricate protein materials, address functional defects or instability in biomedicine, and enhance our understanding of natural protein mechanisms. These interdisciplinary collaborations will pave the way for designing new multifunctional proteins.

1. Introduction

Protein materials are a crucial branch of biomaterial science, attracting significant attention due to their exceptional biocompatibility, bioactivity, and customizable functionality. They show great promise in tissue engineering, biosensing, drug delivery, and biomedicine. Current research primarily focuses on enhancing the mechanical properties, stability, and bio-functionality of these materials. Through employing molecular-level design techniques, researchers can precisely manipulate protein sequences and structures to achieve specific material properties. Using a limited set of side chains and covalent groups, scientists have developed advanced capabilities for creating diverse protein-based materials. Recent progress in protein science and engineering, such as genetic modification, has further enabled the development of biomaterials with enhanced interactions with cells or improved stability within organisms. Complex multifunctional protein materials can also be prepared by integrating proteins with other biocompatible substances. Nevertheless, protein materials face challenges, such as instability under physiological conditions, leading to denaturation, and potential immune responses due to excessive biological activity. Additionally, even modern computational models provide insights into protein structure and function, they still cannot fully predict the three-dimensional structure or behavior in complex systems.

Notably, bottom-up ‘biomanufacturing’ technology skillfully leverages multi-scale self-assembly and supramolecular chemistry, enabled by the versatility of the 20 amino acids in protein and its post-translational modifications (PTMs), for the synthesis of diverse biopolymers. Due to the diversity and arrangement of amino acid residues in each species, proteins exhibit a wide array of chemical functionalities with varying nucleophilicity, pKa values, and redox potentials. The complex and abundant binding sites of proteins enable researchers to optimize and design functional protein materials, efficiently regulating their orientation within these materials. Generally, this optimized structure is achieved by chemically attaching reactive functional groups to proteins, allowing them to interact statically or dynamically with the materials. Moreover, the de novo design, distinct from chemical modification, can unveil functions such as bioactivity and programmed degradability. Consequently, the reaction conditions for protein modification are critical considerations in functionalization strategies [1].

In recent years, researchers have focused on exploring various sources of protein applications, including common proteins/peptides and post-translationally modified (PTM) proteins/peptides. In Fig. 1, this review adopts a stepwise approach, beginning with the fundamental types of protein materials, followed by their applications to identify key challenges, and concluding with strategies for molecular design and optimization to address these challenges. In conclusion, we discuss the challenges and prospects of applying AI to the design of complex protein structures, which could enhance properties such as thermal stability, functionality, and degradability.

Fig. 1.

The schematic illustration of the protein-based materials. The full process of bio-materials from design and structural optimization to application and monitoring, based on AI deep learning, highlights their broad applications and future potential in the fields of medicine, environment, and agriculture.

2. Classification and properties of proteins

2.1. Fibrous protein

Fibrous proteins, such as collagen, silk, and keratin, are characterized by highly repetitive amino acid sequences that endow these proteins with unique mechanical and structural properties in Fig. 2. These repeated amino acid sequences lead to the formation of relatively uniform secondary structures, such as β-sheets, curls, or triple helices, which in turn facilitate the spontaneous polymerization of protein monomers.

Fig. 2.

The protein-based materials of types and source. Fibrous proteins, including collagen, keratin, and silk proteins, provide strong mechanical support; adhesive proteins, such as mussel foot proteins and polydopamine, adhere strongly to various surfaces; elastin proteins have “stretch-relax” capabilities, giving tissues like skin and blood vessels elasticity and flexibility.

2.1.1. Collagen

Collagen, the most abundant protein in the body and a major component of the extracellular matrix (ECM), provides tensile strength to tissues. Its characteristic triple-helix structure, formed by the Gly-X-Y sequence (where X and Y are often proline and hydroxyproline), consists of three α chains with a molecular weight of ∼300 ​kDa [2]. Different collagen types, defined by α chain composition, serve roles in cell adhesion, tissue repair, and tumor suppression, with Type I collagen suited for bone and skin, and Type II collagen for cartilage elasticity [3]. Collagen is primarily sourced from animals, raising concerns about immunogenicity and disease transmission [2]. Recombinant human collagen, produced via genetically engineered microbes, animals, or plants, offers a safer alternative, minimizing batch variability and immune risks [4]. Significant progress has been made in collagen research, and as its applications expand, effective functionalization will be crucial for advancing future collagen materials.

2.1.2. Silk proteins

Filament proteins are composed of repetitive modular units linked by peptide bonds and often feature disordered regions. For example, silk proteins exhibit exceptional mechanical properties, with strengths up to 1.65 ​GPa [5]. Spider silk proteins, such as MaSp1 and MaSp2, feature both ordered and disordered regions forming β-folds [6], while silkworm silk primarily consists of fibroin with a Gly-Ser-Gly-Ala-Gly-Ala sequence, imparting high mechanical strength [7,8]. The β-pleated crystal structure of silk, determined in 1955 using X-ray diffraction, highlights its correlation with properties like adhesion, strength, and elasticity [9]. Silk proteins are sourced naturally from insects or synthetically produced via genetic engineering in microbes, plants, and animals, enabling precise control of their sequence and structure [10,11]. Advances in recombinant silk production address scalability challenges, paving the way for applications in biomedicine, biodegradable materials, biosensors, and high-performance fibers.

2.1.3. Keratin

Keratin-based biomaterials have been extensively studied over the past few decades for their inherent biological properties and excellent biocompatibility. Keratin is classified as hard or soft based on cysteine sulfur content and exists in α-helix and β-fold structures, providing elasticity and tensile strength, respectively [12]. Its mechanical properties are influenced by the orientation and volume of intermediate filaments (IFs). Keratin extraction, involving physical, chemical, or biological methods, is challenging due to its poor solubility and sensitivity to high temperatures, which can degrade amino acids. Functional groups such as carboxyl, amine, and mercaptan allow for chemical modifications [13]. Recombinant methods have enabled high-purity keratin production for structural studies and in vitro filament assembly [14]. Applications include wound healing, drug delivery, and tissue engineering, though challenges remain in achieving correct disulfide bond formation in microbial systems to preserve mechanical integrity [15].

2.2. Adhesive proteins

Proteins secreted by organisms like mussels and sandcastle worms are known for their strong adhesion under various conditions [[16], [17], [18], [19]]. In Fig. 2, many of these proteins, especially mucins, are highly glycosylated, which increases their molecular weight and enables gel-like structures that enhance viscosity in water [20]. The presence of 3,4-dihydroxy-L-phenylalanine (DOPA) is key to their adhesive properties. Some proteins, like elastin, exhibit adhesion even without glycosylation. Elastin-like polypeptides (ELPs) form two phases—dense and dilute—depending on molecular weight, with their transition temperature influencing viscosity and coagulation [21]. In essence, Mussel foot proteins (Mfps) owe their strong adhesion to catechol groups, which undergo redox reactions to switch between oxidized and non-oxidized forms, enabling interactions with metals and organic substrates [22,23]. Recombinant production of Mfps has been explored to reduce extraction costs, though challenges such as aggregation, which impairs adhesive properties, remain.

2.3. Elastin

Elastin, found in connective tissues like skin, lungs, and blood vessels, plays a crucial role in providing elasticity [17]. In insects like flies and dragonflies, resilin in the wing hinge is highly elastic and disordered. Elastin consists of random helices cross-linked into three-dimensional networks through side-chain interactions. Its precursor, tropoelastin, contains alternating hydrophobic and cross-linked blocks, with sequences like [GGVP], [PGVGV], and [PGVGVA]. Cross-linking occurs via lysine residues, forming elastic fibers [24]. Recombinant elastin-like polypeptides (ELPs) mimic natural elastin, maintaining key properties like lower critical solution temperature (LCST) phase behavior [[25], [26], [27]]. Recombinant elastin derivatives (ELRs) offer advantages over natural elastin by enhancing elasticity and mechanical properties through controlled cross-linking, making them ideal for tissue engineering and biomedical applications.

Understanding the fundamental types of protein materials lays the foundation for exploring their practical applications across various fields, as detailed in the following section.

3. Applications of protein-based materials

Building on the understanding of protein material types, this section discusses their real-world applicationsas shown in Fig. 3, highlighting the potential and limitations that drive the need for optimization strategies.

Fig. 3.

Application area of protein-based materials. The wide applications of biomaterials across various fields, including the food industry, biomedicine, and environmental management. In the food industry, biomaterials can be used in functional foods, food additives, and packaging. In biomedicine, they are applied in tissue engineering, diagnosis, and treatment. In the environmental field, biomaterials are utilized for pollution treatment, environmental monitoring, and soil health maintenance.

3.1. Food industry

3.1.1. Additives

Proteins serve as emulsifiers, thickeners, gel-forming agents, and foam agents due to their hydrophobic and hydrophilic regions, reducing interfacial tension in food processing [28]. For example, soy protein isolate (SPI) nanoparticles, prepared by a non-thermal method, act as effective Pickering stabilizers. These nanoparticles are formed through Ca²⁺-induced aggregation and cross-linking with glutaraldehyde, improving emulsion stability by increasing particle size and surface charge [29]. However, natural proteins can face challenges such as low water solubility and reduced emulsification and foaming capacities under certain conditions, including high ionic strength, pH, and temperature, which limits their application in the food industry. The functional properties of food proteins, influenced by processing treatments, affect their suitability as emulsifiers [30,31].

3.1.2. Packaging materials

Proteins are widely used in the food industry for packaging materials, as they help prevent contamination during production, distribution, and storage. These packaging materials protect against moisture, microorganisms, gases, odors, and dust [32]. Protein-based films can be engineered for excellent mechanical and barrier properties through treatments like thermal denaturation, chemical hydrolysis, enzyme treatment, and crosslinking. However, these films may become unstable and lose mechanical strength due to water content and prolonged storage, reducing their barrier function. Current research focuses on improving protein packaging performance through chemical and physical modifications, composite material development, and nanotechnology. Researchers are also exploring functional components integrated into proteins via covalent bonding, grafting, or hydrogen bonding [33]. A new approach involves designing proteins with specific functions based on natural structures to address existing challenges and enhance understanding.

3.2. Environmental applications

3.2.1. Monitoring

Recent advancements have led to the development of protein-nanomaterial hybrids (PN hybrids) for biosensing, combining active proteins (such as enzymes or antibodies) with nanomaterials to enhance biosensor performance. This integration leverages the properties of nanomaterials and the functionality of proteins. While small proteins can be effectively integrated with nanomaterials, binding larger proteins remains a challenge. Bioelectronic ammonia sensors based on protein nanowires from Geobacter sulfreducens demonstrate high sensitivity to ammonia, making them suitable for industrial and environmental applications. Additionally, a supramolecular bio-nanocomposite made from silk fibroin and carbon nanoparticles detects nitroaromatic explosive vapors with excellent response time and reversibility [34]. Despite their potential, protein-based biosensors face challenges in maintaining performance under varying environmental conditions such as temperature, humidity, and pH.

3.2.2. Treatment

Filament proteins with strong polar functional groups—such as hydroxyl, carboxyl, and amino groups—are highly water-soluble and effectively bind metals. Researchers have developed organic-inorganic hybrid structures using a green, one-pot co-precipitation method, resulting in materials with excellent adsorption properties for heavy metal ions in wastewater [35]. Protein materials can also be utilized in environmental pollution treatment through simple techniques like freeze-drying to create aerogels for metal adsorption [36]. Furthermore, they can serve as soil amendments to remediate soil pollution [37]. Although keratin, rich in nitrogen and sulfur, provides nutrients to the soil, its application is limited by its high cysteine content, which resists microbial degradation. Genetic engineering can help reduce cysteine levels or the formation of disulfide bonds, enhancing microbial degradation rates in soil. Additionally, modifying the hydrophilic and hydrophobic regions of protein materials can improve soil water retention, expanding their applications beyond mere soil amendments.

3.3. Biomedicine

3.3.1. Tissue engineering and regenerative medicine

Since the 1990s, Tissue engineering (TE) has gained significant attention and investment in clinical research. Biomaterials like collagen and fibroin serve as scaffolds, supporting tissue repair and regeneration [38]. Protein-based scaffolds are essential for bone tissue repair, which has natural regenerative capacity but struggles with critical defects. These materials must be economical, non-toxic, and biocompatible, with properties that promote bone conductivity and induction [39]. For soft tissue injuries like tendon damage, collagen scaffolds support tendon sheath formation and recruit stem cells for repair [40]. In skin wound healing, protein-based materials, such as gels [41], patches [42], sutures [42], and sponges [43] are used for enhanced repair and healing. Proteins in tissue engineering are versatile and used for scaffolding and regulating biological functions. Through design and engineering, proteins can support cell growth, guide differentiation, and promote angiogenesis.

3.3.2. Drug delivery system

Proteins are widely used in drug delivery systems as carriers, enhancing drug specificity, stability, and interactions. Natural proteins, like serum proteins, protect drugs, prolong circulation time, and increase target tissue concentrations, making them effective in chemotherapy and anti-inflammatory treatments [44]. Proteins can be formulated into nanoparticles to improve drug solubility, extend circulation, and facilitate passive targeting. For example, peptide-based nanoparticles bind hydrophobic anticancer drugs and deliver them to target tissues using mild hyperthermia [45]. Surface modifications and functionalization can improve drug binding and release, while targeting antibodies enable selective binding to target cells for precise drug delivery [46]. Future research should focus on optimizing these modifications, improving drug loading and release technologies, and developing biocompatible, low-immunogenic materials.

3.3.3. Biosensor

Proteins are ideal for biosensors due to their biocompatibility and bioactivity, making them effective as recognition elements and signal transducers in food detection, environmental monitoring, and health diagnostics. For example, silk protein has been used to develop non-invasive electrochemical sensors that mimic natural tissue structures, enabling long-lasting, irritation-free skin contact for glucose monitoring via sweat analysis [47]. The advent of electronic skin highlights the potential of protein-based materials, valued for their flexibility, biocompatibility, and electrical conductivity [48]. Antibodies act as recognition elements, generating signals upon binding to target molecules, while enzymes convert biochemical reactions into detectable outputs for sensitive biomolecule detection [49]. However, improving sensitivity to detect low biomarker concentrations, especially in early disease diagnosis, remains a key challenge in advancing protein-based biosensor technology.

3.3.4. Biomedical imaging

Proteins enhance imaging by improving contrast and sensitivity in detecting disease markers. Imaging probes such as fluorescent dyes, radioisotopes, and magnetic nanoparticles are incorporated into proteins for specific and real-time visualization. Fluorescently labeled proteins are widely used in fluorescence microscopy for non-invasive imaging, while proteins with magnetic modifications improve contrast in MRI for better visualization of tissue structure and function [50,51]. Dual-modality imaging probes, such as albumin aggregates combining fluorescence and MRI, offer high biocompatibility and biosafety. However, challenges remain, including nonspecific binding and reduced binding affinity due to structural modifications, which affect imaging accuracy. Developing proteins with enhanced targeting and stability in biological imaging remains a critical focus.

The practical applications of protein-based materials reveal key challenges that can be addressed through advanced molecular design and chemical modification. Protein structure not only determines drug binding sites and affinity but also directly affects the mechanical properties, stability, and biocompatibility of materials. Conformational, dynamic, and chemical modifications of proteins significantly impact the in vivo behavior of biomaterials, influencing biocompatibility, degradation rate, and biosafety. Therefore, precise modification and optimization of protein structures are crucial for developing functional protein materials for various applications.

4. Strategies for structural optimization of protein materials

4.1. Chemical modification

Protein chemical modification involves altering the structure and function of protein-based materials by introducing various chemical groups into protein molecules or by changing their specific chemical properties. And these modifications can target amino acid residues within the protein, influencing factors such as stability, affinity, and activity. By customizing protein properties in this way, scientists design protein-based materials to meet specific application requirements, as shown in Table 1. Furthermore, coupling proteins to other small molecules or biological macromolecules through chemical modification has significantly advanced the field of chemical biology.

Table 1.

Chemical modification types for various applications, along with their desirable properties.

Method	Molecules	Reagent	Improvement	Application	Refs.
Conjugation	Fluorescent dyes Drugs Polymers	NHS	Peptide functionalization	Functional bioconjugates	[52]
		PEG	Thermo-dynamic stability	Protein-PEG conjugates	[53]
Amino acid modification	Lysine	Phosphorylation	Cellular signaling pathways regulation	Supramolecular hydrogel	[55,56]
	Threonine Tyrosine		Protein functions
	Lysine	Methylation	Cell function maintaining	Single-dose hydrogel patch	[57]
	Aspartic acid Arginine		Environment homeostasis adaption
	Amino acid residues	Glycosylation	Stability	Nanoparticle	[58]
Cross-linking	Amino acid residues Materials	GA	Mechanical strength Stability	Scaffold	[[59], [60], [61], [62], [63]]
		EDC-NHS		Gel
		Genipin		Film
Grafting	Polymers (PDMS, PHB, PLA etc.)	Epoxy saline coupling agent Ferritin Macroinitiator	Anti-FBR effect	Implantable biomedical devices Nanocages	[64,65]
			Functionalization

4.1.1. Conjugation strategies for protein functionalization

Protein coupling involves the conjugation of proteins with various molecules, including fluorescent dyes, drugs, and polymers, to impart new properties or functionalities to the resulting protein-based materials. Protein coupling can encompass interactions between proteins, proteins and polymers, or proteins and small molecules, leading to the formation of novel polymeric structures. The structure of protein contains the sulfhydryl residue, such as cysteine residue, sulfhydryl-acrylamide reaction can be utilized to form covalent bond. This reaction facilitates the incorporation of an acrylamide group into the sulfhydryl residue of the protein, creating a stable covalent bond. Bifunctional reagents such as N-hydroxysuccinimide (NHS)-activated esters can be easily prepared for diverse and complex payloads, facilitating the late-stage bio-conjugation of proteins [52]. The pegylation has been employed for over three decades to enhance the shelf life and pharmacokinetic properties of proteins [53]. This modification technique not only improves the thermal stability of proteins but also extends their circulation time and efficacy. The process involves the use of NHS-activated ester PEG reagents, which feature an NHS group at one end of the PEG chain. This NHS group reacts with the amino groups on the protein, forming a covalent amide bond and incorporating the PEG chain into the protein molecules. This modification alters the surface properties of the protein, enhancing its stability, increasing its water solubility, and reducing its immunogenicity. These improvements are highly advantageous for optimizing the pharmacokinetic and pharmacological properties of proteins. Consequently, the pegylation is commonly utilized in the development of drug delivery systems and biological products, significantly enhancing their performance and effectiveness.

Researchers are increasingly utilizing click chemistry for protein modification due to its precision and selectivity in targeting specific functional groups within complex biological environments. This method is essential for proteins, which often contain multiple functional groups that require selective alterations to maintain biological activity. Click chemistry operates under mild conditions, preserving protein functionality by avoiding harsh temperatures or pH. It also provides various options for adding functional groups like fluorescent markers, polymers, or small molecule drugs, broadening the use of proteins in biomedical research, drug delivery, and biomaterials. A key advantage of click chemistry is its ability to perform certain reactions, especially metal-free ones, inside living cells. This allows for in situ protein modification, aiding the study of protein functions and interactions in vivo. One example is the protein Diels-Alder reaction, where a diene structure is introduced to a protein (e.g., human serum protein). The Diels-Alder reaction occurs between this diene and a nucleophilic reactant, forming a cycloaddition product that covalently anchors the reactant to the protein. This targeted modification can significantly alter the protein's properties and functions, enabling precise functionalization [54].

4.1.2. Amino acid modifications strategies in protein optimization

Protein amino acid modification, studied since the mid-20th century, can be categorized into enzyme-catalyzed and non-enzymatic types. Enzyme-catalyzed modifications involve specific enzymes like kinases, while non-enzymatic modifications rely on chemical reagents to alter amino acids. Advances in biochemistry, bioinformatics, and molecular biology, particularly through mass spectrometry and innovative methods, have made this a key area in molecular biology and biomedicine. Protein amino acid modifications, such as phosphorylation, methylation, glycosylation, and acetylation, introduce functional groups that can significantly alter protein properties and functions. These developments are critical for optimizing protein-based materials, expanding their applications and improving performance.

Hosphorylation primarily targets serine, threonine, and tyrosine residues in proteins. This modification can be introduced through various methods, including enzymatic processes, optical techniques, or chemical reagents such as phosphate acylation reagents and nucleophiles. Phosphorylation is essential for regulating cellular signaling pathways and protein functions, influencing critical processes such as cell division, growth, and apoptosis [55]. Abnormal protein phosphorylation is linked to numerous diseases, including cancer and neurological disorders. For example, a kinase-reactive supramolecular hydrogel developed has been demonstrated potential therapeutic applications. This AURKB-reactive hydrogel releases AZD, an AURKB inhibitor, which effectively downregulates the phosphorylated histone H3 (pH3). The resulting reduction in pH3 expression induces growth arrest and apoptosis in cervical cancer cells, highlighting the hydrogel's efficacy in targeting aberrant phosphorylation pathways [56]. Methylation, a crucial amino acid modification, predominantly affects lysine, aspartic acid, and arginine residues. It plays a vital role in regulating gene expression, protein interactions, and cell signaling, thus maintaining cellular function and homeostasis. In protein biomaterials, methylation can reduce immunogenicity, enhance biocompatibility, and prolong stability in vivo. Additionally, it influences protein interactions, spatial distribution, and assembly within biomaterials. However, abnormal methylation is linked to various diseases, including cancer. For instance, researchers developed a tumor microenvironment-responsive hydrogel patch designed to modulate T-IC plasticity in triple-negative breast cancer (TNBC). This single-dose hydrogel patch, which incorporates an LSD1 inhibitor and a chemotherapeutic agent, effectively inhibits tumor growth, postoperative recurrence, and metastasis [57]. Glycosylation involves the covalent attachment of glycan molecules to amino acid residues in proteins [58]. Glycosylation significantly influences protein folding, stability, and function. However, it remains a non-templated and heterogeneous process, largely due to the involvement of multiple enzymes. Current research is focused on unraveling the mechanisms of glycosylation, its physiological functions, and its associations with diseases. Additionally, efforts are being made to design glycosylated biomaterials for advanced applications in drug delivery and tissue engineering.

4.1.3. Other chemical modification techniques

In addition to common coupling and amino acid modification techniques, other chemical methods like cross-linking play a vital role in biomaterials. Chemical cross-linking, which forms covalent bonds between proteins or with support materials, is often used to prepare protein complexes or immobilize proteins. For example, glutaraldehyde cross-links lysine residues with aldehyde groups, and EDC-NHS is used for collagen-gelatin hydrogels to create biocompatible, porous structures that support cell growth [[59], [60], [61], [62]]. Genipin, a less toxic alternative, reacts with lysine and arginine, forming stable cross-links and secondary amide bonds. Enzymatic cross-linking, such as with microbial transglutaminase (MTG), enhances collagen structure and improves physical properties of protein-based materials [63].

Protein grafting is another key modification technique, involving incorporating polymers or functional groups to enhance material properties. For instance, human serum albumin grafted onto PDMS implants reduced foreign body reactions, improving biocompatibility [64]. Similarly, polyMPC and polyPEGMA were grafted onto ferritin nanocages via ATRP, creating nanostructured materials with advanced functional properties. This technique facilitated the fabrication of microphase-separated nanostructured materials, highlighting the versatility of protein grafting in the development of advanced biomaterials with customized structural and functional properties [65].

Although numerous methods exist for modifying and optimizing protein structures to enhance their performance, achieving high selectivity for specific amino acid residues remains a significant challenge. Many modification techniques can affect multiple sites within a protein, potentially leading to undesirable side effects. Additionally, some methods may require prolonged reaction times or high substrate concentrations, which can compromise the overall efficiency of the modification process. Consequently, future research will focus on developing highly selective modification strategies that precisely target specific amino acid residues while minimizing non-specific effects. Advances in this area are anticipated to enhance the molecular design and modification of protein-based materials, ensuring they meet specific application requirements with reduced unintended consequences.

4.2. Protein molecular design: Optimization and de novo approaches

In the previous discussion, we emphasized the need for selective amino acid modification to overcome the limitations of current protein optimization techniques. In Table 2, large-scale atomic simulations, multiscale modeling, theoretical analysis, and experimental validation offer a comprehensive framework for investigating and improving protein-based materials. These approaches enable precise structural modifications, leading to better protein optimization [66]. Molecular design in science and engineering aims to achieve specific properties by tailoring protein structures and functions through computational and experimental methods. This integration of molecular design and chemical modification advances the development of protein-based materials with desired attributes [67]. Strategies for protein engineering, whether pre- or post-modification, enable researchers to fine-tune protein properties for various applications. Designed strategies are based on an understanding of protein structure, including amino acid sequence, binding sites, and folding. Researchers tailor their approach to the specific goals, whether it involves mutating active sites, introducing functional groups, or optimizing secondary structures. Researchers have synthesized rosiglitazone-based heterodimers with improved α-amylase and antioxidant activities, showing significant in vivo antidiabetic effects [68]. Enhancing biological activity through active site modification or side chain optimization is a key strategy in protein design [69]. In addition to modifying the active site to improve catalytic activity, optimizing the side chains of amino acids can also enhance performance [70]. In catalysis, atomic-level coordination of side chains, substrates, and cofactors is required, and small molecule binding proteins can be designed with flexible binding regions that favor ligand binding in the desired geometry. The results illustrate that unified core packaging and binding site definition is a central principle in ligand-binding protein design. Through precise molecular design and modification strategies, researchers can enhance the functionality and applicability of protein-based materials across various fields [71].

Table 2.

Molecular design of protein-based materials with their functions.

	Method	Model	Target	Functions	Refs.
Protein structure modeling	Homology modeling	Sybyl	Protein Micromolecule	Structural prediction	[76]
				Molecular docking
				Force field calculation
				Drug design
		Modeller	Protein	Structural prediction	[72,86]
				Molecular docking
		SWISS-MODEL	Protein	Structural prediction	[90]
		Discovery Studio	Protein Micromolecule	Molecular modeling	[86]
				Drug design
	Fold recognition	FFAS	Protein	Protein sequence and structure comparison Functional prediction	[90]
		HHsearch	Protein	Domain prediction	[79]
				Unknown protein sequence annotation
		Raptor X	Protein	Three-dimensional structure prediction of a given protein sequence	[91]
		Spark-X	Protein Micromolecule DNA	Fold prediction	[66]
	Deep learning	RFdiffusionAA	Protein Micromolecule DNA	Protein-micromolecule and new proteins complex structure prediction	[81]
				Covalent modifications Prediction
		AlphaFold2	Protein	Protein 3D structure prediction	[78,80]
				Interaction pattern detection between amino acids
		DeepChem	Protein Micromolecule	Small molecule drug prediction	[75,76]
				Molecular property prediction
Protein molecular dynamics simulations		CHARM	Protein Small organic molecule DNA complex	Multi-scale simulations (atomic-level, coarse-grained, and hybrid quantum/classical mechanical)	[69,85]
		AMBER	Protein DNA complex	Quantum mechanical protein folding	[79,80,85]
				Model interactions
				Drug design
				Molecular docking
		GROMACS	Carbohydrates Protein DNA Biomacromolecule complex	Complex biophysical process simulations	[82,[86], [87], [88]]
				Parallel computing capability
		NAMD	Lipid dilayers	Structure and dynamic behavior of the simulated system demonstration	[69,85,86]
			Protein
			Small organic molecule
			DNA complex

While many successes have been achieved in de novo protein design, typically in such problems the sequence and exact backbone structure are unknown. This approach usually involves predicting, designing, and simulating protein structures on computers, followed by experimental validation to confirm the structure and properties of the newly designed protein. Much de novo protein design work has emphasized creating ideal protein structures with unkinked α-helices and β-strands and minimal loops. David Baker developed the protein structure design algorithm Rosetta, which has been used to design numerous novel proteins. The significance of protein molecular design lies in advancing scientific research, expanding applications in biology, medicine, and engineering, and providing new ideas and tools for addressing various social problems. Key strategies include binding site design and amino acid side chain optimization, structural stability design and hydrophobic core optimization, de novo design, and functional fusion. Through these strategies, researchers can push the frontier of scientific research and develop innovative solutions for a wide range of applications, enhancing the functionality and applicability of protein-based materials.

4.2.1. Computational methods in protein design

Three-dimensional protein structure can be predicted using computational algorithms, determining the spatial position of each atom in a protein molecule from the amino acid sequence. Structural prediction methods are traditionally divided into template-based modeling (TBM) and template-free modeling (FM), depending on whether homologous structures can be found in the Protein Data Bank (PDB). Homology modeling, a type of TBM, is based on two principles. Firstly, a protein's structure is uniquely determined by its amino acid sequence, meaning its secondary and tertiary structures can, in theory, be deduced from the primary sequence. Secondly, the tertiary structure of proteins is evolutionarily conserved. Many software and servers are available for homology modeling, including Sybyl, Modeller, SWISS-MODEL, and Discovery Studio, as shown in Table 2. These tools enable molecular design and modification of protein-based materials, allowing researchers to predict and optimize protein structures for various applications.

Both homology modeling and fold recognition can be viewed as template-based modeling approaches. The primary difference lies in the extent of their utilization of structure- and atlas-based terms. Fold recognition extensively employs these terms, whereas homology modeling primarily relies on sequence information. In the last decade, numerous fold recognition methods have been developed, including FFAS, FFAS-3D, HHsearch, RaptorX, DescFold, MUSTER, and SPARK-X. Additionally, several web servers have been made freely available to facilitate fold recognition [72]. However, not all proteins can be identified using homology, and if the homology is less than 30 ​%, only de novo modeling can be performed. The ab initio approach is based on a thermodynamic foundation. The native conformation of a protein corresponds to its lowest energy conformation. Therefore, we can directly predict the 3D structure of protein sequences by constructing appropriate energy functions and optimization methods. Popular tools for de novo modeling include I-TASSER and AlphaFold.

Deep learning is another powerful method that uses statistics to build models representing the real world through massive datasets, with neural networks being the primary tools [73,74]. A neural network is a model composed of multiple layers of interconnected neurons that learn complex patterns within data for prediction and classification tasks. In protein design, effectively utilizing large structural data and applying it to molecular design and modification is crucial. Current deep learning methods for characterizing protein data have seen significant improvements, particularly in protein structure prediction, as shwon in Table 2. These advancements provide valuable insights into protein characterization. The core step is to build a deep learning model that integrates massive upstream data with downstream modeling targets. This approach enables more accurate and efficient design and modification of protein-based materials, enhancing their functionality and expanding their application fields.

Since 2016, DeepChem has been used in applications such as inhibitor design modeling for BACE-1 [75] and the development of novel deep learning techniques for drug discovery [76]. When more refined physical simulations and energy functions are required to better account for atomistic-level interactions, tools like RFdiffusion, based on RoseTTAFold, have been developed. RFdiffusion can design functional proteins with atomic precision and has been extended to nucleic acids and protein-nucleic acid complexes [77]. It enables the design of protein-ligand interactions and can create different functional proteins from simple molecular specifications, similar to generating images from user-specified inputs [78]. AlphaFold is another powerful tool for structure prediction with high accuracy. It employs deep learning techniques, including convolutional neural networks (CNN) and residual neural networks (ResNet), to learn the complex relationship between protein sequence and structure [79]. AlphaFold does not depend on specific physical simulations or rules, allowing it to make structural predictions for various protein types. However, predicting large protein complexes remains challenging due to the size and complexity of interactions among multiple subunits [80]. Ben Shor introduced CombFold, a combinatorial and hierarchical assembly algorithm that predicts the structure of large protein complexes based on AlphaFold-Multimer (AFM) predicted input subunit pairs or substructures of larger subunits. AlphaFold and RoseTTAFold provide excellent structural predictions for native proteins because multiple sequence alignments contain substantial coevolutionary and other information that reflects the three-dimensional (3D) structure. However, when only a single sequence is available, the structural model is often less accurate compared to protein complexes. Current efforts are more effective at predicting binding to single-stranded protein structures.

The ProteinMPNN method which requires only a fraction of the time compared to physics-based methods like Rosetta for solving the sequence design problem. ProteinMPNN achieved higher sequence recovery on the native scaffold (52.4 ​% vs. 32.9 ​%) and successfully rescued previously failed designs for protein monomers, assemblies, and protein-protein interfaces that were originally addressed by Rosetta or AlphaFold. Many essential protein functions, such as receptor signaling, immune evasion, and enzymatic activity, involve covalent modification of amino acid side chains with sugars, phosphates, lipids, and other molecules. The RoseTTAFold All-Atom (RFAA) model simulates these modifications by treating residues and chemical parts as individual atoms (with corresponding covalent bonds to atomic tags) and considering the remaining protein structure as residues. Although RFAA is highly useful for designing protein-small molecule binders and modeling complex biomolecular assemblies, its accuracy needs further improvement to significantly impact drug discovery [81]. RFAA and RFdiffusionAA are anticipated to become widely used for modeling and designing complex biomolecular systems, thereby advancing the molecular design and modification of protein-based materials.

4.2.2. Molecular dynamics simulations in protein structure prediction

In protein structure prediction, molecular dynamics (MD) simulations are used to explore potential conformations and identify the lowest energy state by simulating the system's transformation between various conformations, as shown in Table 2. This approach enables researchers to study protein folding by simulating dynamics within or between protein molecules [82]. By analyzing the movement of protein molecules over time, MD simulations help identify more stable conformations and enable the achievement of specific functions [83]. Furthermore, MD simulations can be employed to examine interactions between proteins and the immune system, offering valuable insights into protein behavior in a biological context. This methodology promotes the molecular design and modification of protein materials by offering a deeper understanding of protein stability, function, and interactions [84].

Early applications of molecular dynamics (MD) simulation software primarily used CHARMM, which offers a broad range of functions, including protein folding and interactions with small molecules. Investigation of interactions between biological macromolecules and small molecules facilitates modeling of the system at different levels. Multi-scale simulations, including atomic-level, coarse-grained, and hybrid quantum/classical mechanical simulations, enable a more comprehensive exploration of protein structure, function, and dynamic behavior [85]. To study interactions between proteins and their surrounding environment (such as solvents or ligands), particularly in active site reactions, QM/MM (quantum mechanics/molecular mechanics) simulations are necessary. The combination of high-precision quantum mechanical descriptions with the efficiency of classical mechanics enables the simulation of large biomolecular systems and complex reaction mechanisms. Therefore, CHARMM is often used in combination with another MD simulation software, AMBER. CHARMM handles the classical mechanical aspects, while AMBER focuses on the quantum mechanical components. AMBER is a widely used MD software suite for studying biomolecular structure and dynamics. It provides a variety of tools for protein folding, interaction modeling, and more. AMBER is especially known for its force field parameters and simulation methods, with extensive experience in computational chemistry and drug design, particularly for studying drug-protein interactions [85]. Another MD simulation software, NAMD, incorporates both CHARMM and AMBER force fields and specializes in large-scale MD simulations of biological molecules, particularly large proteins and membrane proteins. NAMD can be integrated with molecular graphics programs through communication systems, offering interactive tools for viewing and modifying simulations in real time [86]. Additionally, GROMACS is one of the most widely used MD simulation software due to its high performance and open-source nature, making it particularly suitable for large-scale simulations. It offers a variety of force fields and advanced algorithms for simulating different types of biomolecules. GROMACS employs Hamiltonian and Newtonian equations to describe particle motion in protein systems, accounting for potential energy, bonded interactions (bonds, angles, dihedral angles), and nonbonded interactions between atoms. An integral algorithm simulates the evolution of proteins over time. Simulation results can be visualized using software such as VMD or PyMOL, aiding in the understanding of protein structure and dynamic behavior [87]. However, GROMACS employs various force fields, each with limitations and potential inaccuracies in describing certain systems or interactions. Therefore, using additional molecular simulation software or tools for result validation is recommended. An automatic perturbation topology builder using a graph-matching algorithm has been developed to identify the maximum common substructure (MCS) of multiple molecules. This tool generates perturbation topologies suitable for free-energy calculations using the GROMOS and GROMACS simulation packages, enhancing the applicability of perturbation free-energy methods for estimating protein stability, binding affinities, and rational drug development [88]. Combining multiple dynamic simulations can provide more comprehensive and accurate information, enhancing the reliability of protein structure and dynamics predictions and accurately revealing processes such as protein folding, unfolding, and ligand binding.

Protein structure is predicted by integrating fractal design and kinetic simulation, however, significant challenges remain. We are far from predicting which drugs will fail in development, so drugs promising in animal models may not pass clinical trials. Additionally, sequence-independent docking software for protein design is underdeveloped, particularly for arbitrary small molecules, remaining a prominent challenge. Despite these challenges, integrating molecular design and kinetic simulation offers a robust framework for advancing the molecular design and modification of protein-based materials. Continued research and development will help overcome current limitations and enhance the predictive power and practical applications of protein engineering.

5. Challenges and prospects

Proteins serve as essential building blocks in nature, forming molecular machines and dynamic materials like cytoskeletal components and muscle fibers. Over the past two decades, their complexity and functionality have inspired efforts to design novel proteins. The structure, properties, and functions of proteins are primarily determined by their amino acid sequence and folding. To better understand these relationships, researchers have combined tools from fields like biophysics, supramolecular chemistry, and materials science. This interdisciplinary approach has expanded the applications of protein biomaterials and advanced protein design techniques. Many innovative methods and software have emerged, enhancing the design and modification of proteins. This review highlights the applications of fibrin, adhesins, and elastin in fields such as food, environment, and biomedicine, while also discussing strategies for optimizing protein structures and progress in rational protein design.

Several challenges remain before the novel proteins discussed in this review can be widely applied. Despite the vast diversity of natural proteins, many biomedical and industrial issues remain unsolved. Scientists are now focusing on using computational analyses to uncover the principles of protein structure and function to design custom proteins with specific functionalities. With the rise of machine learning and advanced computational modeling, the future may involve the creation of entirely new proteins. To ensure cost-effective, efficient, and sustainable production, molecular design will need to integrate high-throughput screening with automated systems. This computerized approach offers a data-driven method for developing protein polymers. In summary, combining computational tools, machine learning, and innovative molecular design holds great potential for overcoming current limitations, paving the way for the development of protein-based materials with customized functionalities across biomedicine, industry, and other fields.

Artificial intelligence (AI) offers new opportunities for designing novel proteins with specific functions. While traditional protein design has focused on regions with well-defined structures, AI-based approaches are challenging this paradigm by enabling the design of biomolecules that rely on flexible regions, allowing for diverse conformations. Machine learning and AI-based computational tools have shown great promise in predicting real-world protein structures. Researchers are designing proteins for sensor applications or to control cellular functions with RFdiffusion. However, designing proteins with complex active sites that require multiple key amino acids remains a challenge for AI systems. Despite these challenges, recent advancements in AI have transformed protein design, allowing designers to generate sequences and structures at unprecedented speeds [89]. However, the performance of AI models is highly dependent on the quality and quantity of training data available. Precise computational methods for designing complex protein targets, such as enhanced enzyme activity or altered ligand specificity, remain elusive and difficult to achieve. To overcome these challenges, collaboration among biologists, computational scientists, materials scientists, and engineers is crucial. Through collaborative efforts, we can advance the development and application of complex protein materials, tackle functional defects or instability in biomedicine, and enhance our understanding of natural protein mechanisms. Such collaboration will facilitate the creation of multifunctional proteins suitable for a broad range of applications.

Author contributions

Conceptualization, S. X.Y. Z., A. T. and Z. Z.; writing—original draft preparation, A. T. and Z. Z.; validation, A. T., Z. Z., and X. R.; writing—review and editing, S. X. and Y. Z.; visualization, A. T. and Z. Z. All authors have read and agreed to the published version of the manuscript.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.