Bridging the Chemical and Functional Divide: The Role of Gene Expression Prediction in Drug Discovery

Freedom Preetham
Meta Multiomics
Published in
6 min readJul 10, 2023

The world of drug discovery is a complex, multifaceted universe filled with vast chemical landscapes and intricate biological mechanisms. The creation of a new drug is an endeavour that often on average spans over 12 years and involves $2.6 billion dollars. This significant commitment of time and resources is necessitated by the scale of potential molecular structures available for synthesis and examination, a theoretical space estimated to contain an astounding 10⁶⁰ molecules.

To understand the magnitude of this figure, it’s useful to contrast it with the known universe, which houses roughly 10⁸⁰ atoms. Essentially, the “chemical space” of drug discovery is a virtually infinite expanse that defies comprehension. Adding to this complexity, the rapidly evolving field of genomics has presented us with an additional task: identifying which genes to up-regulate or down-regulate, a search space of 4^3.2 billion possibilities which is technically 10¹⁹²⁶⁵⁹¹⁹⁷² combinations. The current functional search space within that varies between about 10²¹ to 10¹⁰⁰ genomic targets given the number of genes and it’s genetic variants.

Despite advancements in technology, even the most sophisticated experimental facilities can process merely around 10⁵ compounds daily. Thus, traditional experimental methodologies alone are woefully inadequate to explore the entire chemical space.

Deciphering the Chemical and Functional Spaces

Chemical Space: This term refers to the theoretical domain encompassing all possible chemical structures, regardless of their known existence or synthesis in the laboratory. To give an idea of the vastness of this space, it’s estimated that there are about 10⁶⁰ possible molecules that could be created with just the elements commonly found in drug compounds. Chemical space, therefore, represents the totality of these potential molecules. Drug discovery can be seen as a journey through this space to find regions where useful drugs exist, akin to finding habitable planets in the universe.

Functional Space: Once a potential compound (a point in the chemical space) has been identified, we need to understand what this compound can do biologically, that is, its function. This is where the concept of functional space comes in. The functional space refers to all possible biological activities that a compound can exhibit. This might involve binding to a certain protein, triggering a certain cellular pathway, or impacting the body’s physiology in a particular way. Each compound’s potential effects, beneficial or adverse, are part of its location in the functional space. Drug discovery, in the context of functional space, involves finding compounds that occupy the right areas, those that produce beneficial effects without causing harmful ones.

In practice, it means identifying whether a compound can bind to a target, evaluating its binding efficacy, and determining the resulting effect once bound. Imagine it as finding not just any key that fits into a lock, but the one that turns it in the right direction to unlock a therapeutic benefit.

Pioneering Approaches in Drug Discovery

To navigate these vast and complex spaces, the world of drug discovery has turned to computational techniques. These strategies can be broadly divided into three categories: Simulation, Virtual Screening, and De Novo Drug Design.

  1. Simulation: This method forms the bedrock of in-silico drug discovery, employing techniques such as Quantum Molecular Dynamics and Molecular Docking to simulate and predict molecular interactions based on fundamental physical principles. While incredibly accurate, these methods are also computationally intensive, sometimes requiring days to process a single compound.
  2. Virtual Screening: This faster approach allows researchers to predict various properties of a compound, such as solubility, toxicity, and notably, binding affinity. Capable of processing up to 10⁸ compounds daily, it is typically limited to commercially available compounds.
  3. De Novo Drug Design: The most ambitious of the three, this approach allows scientists to venture beyond known structures. Using evolutionary strategies and Generative AI, it’s possible to solve the inverse problem, i.e., identifying the necessary chemical structure for a given function, thereby reverse-engineering the drug discovery process.

Harnessing Gene Expression Prediction in De Novo Drug Design

One of the most significant leaps in De Novo Drug Design has been the incorporation of gene expression prediction, facilitated by advancements in Generative AI. Gene expression profiling, or determining which genes are being actively transcribed in a cell at a given time, plays a crucial role in understanding cell functions, disease progression, and the potential impact of drug compounds.

Delving Deeper: The Role of Gene Expression Prediction

In the intricate world of drug discovery, gene expression prediction has emerged as a cornerstone of De Novo Drug Design. At its core, gene expression prediction aims to anticipate how various genes will behave under different conditions, offering insights into how a potential drug might affect the cellular environment.

The power of gene expression prediction extends beyond simply determining the effects of a drug. In many ways, it helps us unravel the complex tapestry of biological processes that underpin health and disease, allowing us to predict and understand how diseases evolve and how our bodies respond to various therapeutic interventions.

Generative AI: The Game Changer

Generative AI has transformed the landscape of gene expression prediction. By creating synthetic gene expression data, it enables researchers to predict how a potential drug compound could affect gene expression. This ability to anticipate the impact of a drug on the gene expression profile of a cell has far-reaching implications for the field of drug discovery and personalized medicine.

The predictive ability of Generative AI goes hand in hand with the development of new drugs. As we move away from a one-size-fits-all approach in medicine towards more personalized treatments, the ability to predict how individuals’ gene expressions will respond to certain drugs is invaluable.

Moreover, as our understanding of genetics grows, we are beginning to realize that many diseases, such as cancer and neurodegenerative disorders, are influenced by the complex interplay of numerous genes rather than being caused by single-gene mutations. Understanding these gene networks and being able to predict how they change in response to drugs can help us design more effective treatments.

Industry Example: Cognit.AI in De Novo Drug Discovery

Cognit.AI is a generative AI in genomics startup in San Francisco, which is advancing gene expression prediction and engineering, creating significant implications for the de novo drug discovery process. Their innovative platform can analyze any DNA sequence across various cell types and predict gene expression and transcriptomic profiles. Even more impressively, the platform can predict gene expressions influenced by mutations in distal regulatory elements, which could be up to 1Mb away from the gene locus.

In terms of de novo drug discovery, Cognit.AI’s tools could play a revolutionary role. The process of de novo drug design relies on constructing novel molecular structures that can interact beneficially with biological targets. Here’s where gene expression prediction comes into play. By understanding how changes to DNA sequences or regulatory elements might impact gene expression across different cell types, researchers can anticipate the impacts of novel drugs on the body’s genetic machinery.

Moreover, Cognit.AI’s perturbative oracle enables in-silico experiments akin to CRISPR, facilitating rapid gene expression engineering. Along with in-silico saturation mutagenesis experiments, these tools offer unprecedented insights into gene functions, interactions, and impacts of potential drugs. These could help identify promising drug candidates or predict potential side effects, making the drug discovery process more efficient and targeted.

The Convergence of Techniques: Paving the Future of Drug Discovery

The vastness of the chemical and functional spaces poses monumental challenges in drug discovery. However, the integration of computational techniques like Simulation, Virtual Screening, and De Novo Drug Design, particularly with the incorporation of gene expression prediction, promises a future where drug discovery is not only more efficient but also remarkably precise.

While none of these techniques alone offer a magic bullet, their combination, coupled with continual advancements in AI and machine learning, is driving a significant shift in the drug discovery landscape. The future promises a more personalized approach to treatment, shaping an era of medicine that’s accurately tailored to the individual’s genetic makeup, signaling the dawn of truly personalized and precision medicine.

--

--