Target Identification: The Cornerstone of the Drug Discovery Process

Table of Contents:

1. Selecting the right therapeutic target
2. AI and ML in target identification
3. Large Language Models (LLMs) for target identification
4. Omics data analysis with AI and ML
5. Target ranking using AI and ML tools
6. Ardigen’s approach to target identification
7. Target identification: the first step to successful drug development

Target identification is the cornerstone of developing successful therapies. As the first step of the drug discovery process, it sets the entire pipeline for success. If the drug target is not chosen appropriately, this could lead to significant losses in terms of both time and resources. The time it takes to develop new treatments exceeds 10 years and the average cost to bring a new therapy to market is in the $1-2 billion range [1]. Therefore, you want to be sure that you have selected the right target to go after from the very beginning. 

The importance of the right target identification is evident in the following statistics: nine out of ten drug candidates fail at the clinical stages [2]. According to clinical trials data, the most common reason for failure is lack of clinical efficacy (with 40-50% failure rate), followed by toxicity ( 30%), poor drug-like properties (10-15%), and overestimated commercial potential (10%) [3]. With the lack of clinical efficacy accounting for nearly half the failures, correct initial target selection is crucial for reducing the attrition rate.

Selecting the right therapeutic target

What does correct target identification mean? The most important consideration is that the target must be associated with the underlying disease. The second is that the target must be druggable, which means it can be modulated by a drug, producing a desired therapeutic action. This is where the nuances of target identification come into play. For example, in some cases, a target may be associated with the disease but not essential. In this case, a drug may act on a target but not produce the desired therapeutic effect. In other cases, there may be compensatory molecular mechanisms that yield the therapeutic intervention ineffective.

Experimental approaches to target identification can sometimes generate targets that may turn out to be non-essential during subsequent validation stages. The complexity of biological systems makes it difficult to predict all the potential interactions that can yield a target ineffective. For this reason, artificial intelligence (AI) and machine learning (ML) methods have emerged as valuable complementary approaches, offering new and expanded capabilities for target identification.

AI and ML in target identification

Advanced computational approaches, including AI and ML, have supercharged the drug discovery and development process, including target identification. With the power to uncover patterns and insights lost in conventional data analysis, ML is an indispensable tool in the hands of bioinformaticians, biologists and clinicians who are dedicated to bringing drugs to market faster and with higher success rates.

ML and AI methods can help optimize the signal-to-noise ratio and find actionable patterns in data. For example, one of the challenges in the traditional ‘one target, one drug’ model is the presence of redundant functions and compensatory signaling routes. AI enables complex network analysis and elucidation of compensatory functions that can facilitate the discovery of synthetic lethality mechanisms [4, 5].

Large Language Models (LLMs) for target identification

AI can be used to mine the literature for target selection. Domain-specific large language models (LLMs), like BioGPT from Microsoft [6], can support therapeutic target discovery by quickly mining biomedical texts. Trained on extensive data from millions of publications, these tools can link diseases, genes, and biological processes, enabling the rapid identification of disease mechanisms, potential drug targets, and biomarkers. However, this approach introduces an inherent bias towards mechanisms and targets that have already been previously identified.

Omics data analysis with AI and ML

Multiomic data offers a wealth of information for target identification. In the hands of domain experts such as omics researchers, AI and ML tools can provide powerful ways to uncover novel therapeutically relevant targets, as well as assess their druggability potential. AI and ML approaches can be used for analyzing data from high-throughput CRISPR screenings, as well as genomic, transcriptomic, proteomic and metabolomic studies. It has been demonstrated that AI and ML approaches can reduce the time to identify viable targets from 4-5 years with traditional methods down by 70% [7].

Target ranking using AI and ML tools

AI and ML can also help rank targets according to their therapeutic relevance [8]. The ranking includes different scoring metrics, such as how druggable or clinically relevant the target is, the drug specificity and safety, its novelty and economic potential, and others. The ranking process helps uncover indication-relevant biological patterns, evaluate the safety of your target (based on literature and clinical trial information), as well as predict drug responses in vivo from preclinical data to improve translatability [9].

Ardigen’s approach to target identification

As a leader in AI in drug discovery and development, Ardigen has spent the last nine years creating a framework that facilitates rapid and efficient target identification. Ardigen’s Target ID Services harnesses the power of its own data universe, combined with ML and AI techniques developed specifically to streamline target identification and reduce the risk of the drug discovery and development process.

Our custom ML and AI-driven target identification approaches include curated literature searches, including annotated and unannotated databases, as well as analysis of proprietary biological datasets. ML and AI methods enable comprehensive analysis of multiomic data (including genomic, epigenomic, transcriptomic, proteomic and metabolomic profiling), functional genomics studies and phenotypic profiling image data analysis (such as High Content Screening). 

In addition to target identification, our framework can also aid in the subsequent target validation steps by providing initial in silico validation to find associations with indications and phenotypes. We combine multiple modalities that enable not only a thorough target identification search but also a comprehensive assessment of the target’s viability, developability and therapeutic potential.

Target identification: the first step to successful drug development

Thinking of target identification as the cornerstone of your research program will help you bring effective therapies to market faster, cut down the costs of R&D and reduce the attrition rate. You can elevate your target identification efforts by implementing AI and ML methods, which are essential to staying competitive in the market. 

With Ardigen’s Target ID Services, you can shorten your target identification timelines, ensure the target efficacy, druggability and therapeutic relevance, as well as tackle the initial steps of the target validation process. Discover the capabilities of Ardigen’s Target ID Services today—contact us to learn more. 

Works Cited: 

1. Wouters, Olivier J., et al. “Estimated research and development investment needed to bring a new medicine to market, 2009-2018.” Jama 323.9 (2020): 844-853. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7054832/
2. Sun, D., et. al. “Why 90% of clinical drug development fails and how to improve it?” Acta Pharmaceutica Sinica B 12(7) (2022): 3049-3062. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9293739/
3. Dowden, H., and Munro, J. “Trends in clinical success rates and therapeutic focus.” Nat Rev Drug Discov, 18(7) (2019): 495-496. https://www.nature.com/articles/d41573-019-00074-z
4. Ryan, C. J., et al. “Complex synthetic lethality in cancer.” Nature Genetics, 55(12) (2023): 2039-2048. https://www.nature.com/articles/s41588-023-01557-x
5. Wang, J., et al. “Computational methods, databases and tools for synthetic lethality prediction.” Briefings in Bioinformatics, 23(3) (2023): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9116379/
6. Luo, Renqian, et al. “BioGPT: generative pre-trained transformer for biomedical text generation and mining.” Briefings in Bioinformatics 23.6 (2022): https://academic.oup.com/bib/article/23/6/bbac409/6713511
7. Pun, F. W., et al. “AI-powered therapeutic target discovery.” Trends in Pharmacological Sciences (2023): https://www.cell.com/trends/pharmacological-sciences/fulltext/S0165-6147(23)00137-2
8. Muslu, Özlem, et al. “GuiltyTargets: prioritization of novel therapeutic targets with network representation learning.” IEEE/ACM Transactions on Computational Biology and Bioinformatics 19.1 (2020): 491-500. https://ieeexplore.ieee.org/document/9121705
9. Obrezanova, Olga. “Artificial intelligence for compound pharmacokinetics prediction.” Current Opinion in Structural Biology 79 (2023): 102546. https://www.sciencedirect.com/science/article/abs/pii/S0959440X23000209