How is bioinformatics improving drug discovery?

As pipelines cada vez mais robustas para análise de dados de sequenciamento têm nos permitido detectar, com alta sensibilidade, em testes não invasivos, mutações de resistência e novas fusões gênicas, que até poucos anos atrás não eram conhecidas, e que hoje podem ser detectadas de forma precoce. Além disso, os algoritmos de bioinformática que detectam instabilidade genômica ou aqueles capazes de selecionar pacientes que se beneficiarão do uso da imunoterapia possibilitam que a medicina de precisão chegue a um número cada vez maior de pacientes. É a detecção de novos biomarcadores, através de um sequenciamento amplo, que fornece substrato para o desenvolvimento de novas terapias.

Bioinformatics in the era of AI & ML serves as the bridge that connects the dots in the vast landscape of biological information. It takes the complex genetic and molecular data from various sources and translates it into a language that AI and ML algorithms can comprehend. This translation enables these algorithms to sift through enormous datasets, recognize patterns, and pinpoint potential drug targets by identifying genes, proteins, or pathways associated with diseases. Thus, bioinformatics empowers AI and ML to significantly expediting the process of finding potential drug candidates. This collaborative effort between bioinformatics and advanced technology holds great promise for more efficient and targeted drug discovery.

The growing body of NGS data and metadata is making is making Target ID a very exciting area in bioinformatics. I'll stick to gene expression for this post. Gene expression patterns are very sensitive to cell types, disease states, and treatments. These patterns are very powerful for IDing similarities and differences when a target is perturbed. For examples, does hitting prospective target 1 exacerbate of rescue a disease state? Does the resulting pattern look similar to attributes seen elsewhere. This required validated and clean data and metadata, which is still hard to come by. Bioinformatics is just as key for deciding which datasets should be used, as the insights derived from the data.

Bioinformatics plays a crucial role in improving drug discovery by providing tools and methods to analyze and interpret biological data on a large scale. By integrating genomic and proteomic data, bioinformatics aids in identifying specific molecules or proteins associated with diseases. These molecules can serve as potential drug targets, guiding researchers in the development of targeted therapies.

Bioinformatics and compchem has been pivotal and has contributed to several drug discovery success stories, where computational methods like pharmacophore modelling, docking and AI/ML modelling have led to several lead and clinical candidates' development. For example, Tirofiban (brand name Aggrastat), is an antiplatelet medication and is the first drug candidate whose origins can be traced to a pharmacophore-based virtual screening lead. Other computational methods like, docking, virtual screening, and AI/ML has also played a key role in several success stories. In short, these methods not only pinpoint potential lead candidates but also reduce the time and cost associated with experimental assays, making drug discovery more efficient.

In general, virtual screening (i.e., the use of in-silico methods to search for active compounds) tends to be more successful the more you know about your target. If you are interested in a well-characterized target, you might be able to find known actives in databases such as ChEMBL or PubChem. These compounds could serve as starting points for your search or allow you to derive a pharmacophore model. If you have access to the three-dimensional structure of your target (either from experiments or predictions such as AlphaFold), you can use docking methods. It is important to keep in mind that virtual screening might increase your chances of finding actives, but still only generate hypotheses requiring experimental validation.

A compound needs to fulfill many criteria to qualify as a drug candidate: it needs to be safe and effective, but also convenient to use. Lead optimizations is thus often described as a complex multi-parameter optimization challenge. For many relevant optimization parameters, data from historical projects is available and can be used to train machine learning methods. A detailed mechanistic understanding of the interactions of the compound with its targets and the whole organism can help to assess the trade-offs between different parameters and to define a global objective functions.

One important element is the quality and quantity of data going into Bioinformatics pipelines. This affects downstream utility.

In this context, being able to utilize advanced genomics, data analytics, and AI, is a way to reveal specific patient groups identified through molecular profiling. This process allows for stratified and efficient patient treatment within focused clinical trials, with less participants and potentially better results. For instance, Lantern Pharma’s RADR® AI platform managed to predict Elraglusib’s response to metastatic melanoma, resistant to checkpoint inhibitors, with 88% accuracy--for a group of 26 patients.

In the realm of clinical trials, bioinformatics emerges as an indispensable ally, requiring a collaborative and multifaceted skill set. To excel in this field, one must combine a solid grasp of clinical science, statistics, data analysis, and coding. Successful clinical bioinformaticians employ their expertise to work closely with healthcare professionals and research teams, effectively navigating the intricate landscape of trial data. In the context of clinical trials, the ability to harmonize diverse skill sets and collaborate seamlessly is pivotal to driving innovation and uncovering transformative insights for patient care and medical advancements.

One notable example demonstrating the power of bioinformatics and AI in drug repurposing is a well-known case of baricitinib. Back in 2020, when the COVID-19 pandemic struck, UK-based company BenevolentAI used its Knowledge Graph computational system to rapidly, within days, repurpos baricitinib, a known treatment of rheumatoid arthritis, to as an effective medication against the cytokine storm associated with severe cases of COVID-19. By swiftly analyzing vast datasets of medical information, the AI identified the drug's potential efficacy in interrupting the pathways of the novel coronavirus. Later, the effect was confirmed in clinical trials and in 2022, FDA granted Eli Lilly and and Incyte approval.

USP of Bioinformatics in Drug Design is its key role in Drug Discovery pipeline with time and cost reduction. Though the usage of the word "Novel drug" in academic researches drastically increased with the application of bioinformatics tools and approaches, the commercial research uses it cautiously. Drug innovation or novel drug design is mostly happening at the lead optimization stage though the several innovative approaches evolved for the different stages of drug design. Accessibility to the patent, literature and compound databases and its effective curing and comparison with compound in development help to assess whether the compound is novel or no easily.

One of the classic case of Covid-19 vaccines. Bioinformation via AI accelerated the development of the Covid-19 vaccines. Ordinarily it takes around 7 to 10 years to develop a vaccine. Using AI and bioinformatics Covid-19 vaccines were developed in less than a year. AI and bioinformation has been used for the good of humanity.😊💐