De novo protein design—From new structures to programmable functions

Background

The field of protein design is undergoing a fundamental and practical change: instead of modifying existing proteins, it is now possible to build from scratch proteins with complex structures and functions that are just as powerful as those found in nature, but new and programmable. The emergence of Artificial Intelligence (AI) has greatly facilitated the de novo design of proteins and is changing the way design is conceptualized.

Discussion

Concepts and Methods for De Novo Protein Design

While traditional approaches to protein design have relied on physical principles and atomic-level representations, advances in AI have rapidly revolutionized the field and addressed many key design challenges. Instead of starting from evolved proteins that already exist in nature, novel, ab initio protein design aims to extend the space of protein structures, sequences, and functions beyond what is known in nature.

Figure 2. Protein design concepts and approaches.

Highlighting these advances, key challenges, and future opportunities, Tanja Kortemme of the University of California, San Francisco, published a research review, De novo protein design-From new structures to programmable functions, in the journal Cell. The review introduces the concepts and methods of de novo protein design, and then discusses (1) the frontiers of new protein structure design, (2) new molecular functions, (3) interfaces between "de novo" proteins and cellular functions, and (4) perspectives on long term and emerging issues, with an emphasis on the last five years of advances in protein de novo design.

Table 1. Recent computational protein design reviews with title, short summary, and reference

De novo design of molecular functions

Typically, computationally designed proteins provide a starting point where activity can be reliably measured but is low, and subsequently needs to be experimentally optimized to achieve a practically useful function. With advances in deep learning methods, this paradigm is beginning to change, at least for the initial set of functions. The process of computationally designing a function typically involves two steps: the first defines the requirements for the function, and the second optimizes the protein structure and sequence to match those requirements. As the use of deep learning in protein design continues to advance, the way these steps are performed is changing rapidly and achieving significant success rates.

Figure 3. De novo design of molecular functions.

De novo design for cellular function

The application of protein de novo design to cellular functions is opening a new era of synthetic signaling systems that have important applications in fields such as basic biology, bioengineering, and medicine. In addition, this approach may allow us to access new functions that are not yet available in nature. In terms of cellular function, advances in computational methods allow us to design sensors and actuators with diverse inputs and tunable outputs. These systems are capable of sensing and responding to molecular signals and translating these signals into output responses within the cell.

Figure 4. De novo design to control cellular functions

Conclusions

With the rapid development of new methods, successful experimental applications are now focused on relatively simple problems such as the design of idealized folds, symmetric assembly and protein-protein interfaces. The increasing success rate of these applications has made some long-standing challenges, such as the design of precise geometries for polar functional sites and the design of dynamic proteins, more achievable. Recent developments, such as protein diffusion modeling, can be used to generate proteins from scratch around small-molecule ligands, although screening of a large number of designs is still required. Further design goals, such as molecular machines, are also becoming accessible, and more complex complex functions can be deconstructed into designable components and ultimately realized.

Related Service We Offer