Bioinformatics & Biotechnology AAP

Brandon W. Higgs

In the biopharmaceutical industry, virtually every scientific and technologic advancement is facilitated by the collection of data. The sheer volume of information has necessitated the development of an entirely new discipline —bioinformatics —to integrate, analyze, store and interpret data to improve our understanding of biological processes. In the last decade, the field of bioinformatics has exploded in response to the evolution of digital imaging modalities, electronic medical records, high-throughput DNA/RNA sequencing, cloud computing, and wearable health devices. Bioinformatics has been applied to address these new data forms, most notably with advancements in artificial intelligence (AI).

AI has become a popular means to address a gamut of problems across industries. Within pharma and biotech, AI has been touted as delivering the promise of precision medicine, by identifying the right drug at the right dose for the right patient. This branch of computer science is certainly not new to the bioinformatics field, as AI has grown from the early seeds of statistics, evolutionary computation, neural networks, and machine learning to the more autonomous algorithms used today. However, these approaches have been of limited value in the past, primarily due to both computational and data constraints. Multi-modal data from large patient cohorts are now keeping pace with algorithmic complexity and previously deemed hyper-dimensional datasets are now considered modest and manageable. Further, new applications for AI engines are continuously being identified. AI has begun to complement, or in some cases, replace manual approaches across the pharma and biotech industry. AI-driven digital pathology has improved characterization of the immune landscape of a tumor, as compared to a pathologist’s manual assessment under a microscope. Radiographic image features can be more resolved with ‘learned’ segmentation strategies, compared to a physician’s evaluation of tissue abnormalities. Automated AI-engines can ‘learn’ relevant clinical characteristics to extractand derive a patient’s medical history from an EHR. At the molecular level, AI has been used to model the kinetics between a patient-specific cellular receptor and its cognate ligand to better understand how to induce T cells to trigger an immune response.

With these advancements come challenges such as storage, computational complexity, security and privacy, and integration to drive meaningful results and ultimately improve human health and well-being. AI is reshaping the way patients are diagnosed, treated, and monitored at a rapid pace, though the complexity and underlying assumptions behind these algorithms can often result in inappropriate implementation. As with many methods used in bioinformatics, the traditional proverb holds that garbage in produces garbage out. Properties of skewed dimensionality, sparse data matrices, and inherent complexity of modeling biological systems can drive distorted solutions, suboptimal convergence, and massive overfitting, thus highlighting the importance of identifying appropriate applications for AI. Nonetheless, bioinformatics has begun to leverage AI as a fundamental technique in its copious repertoire of tools to address a multitude of scientific problems.

"Bioinformatics is a dynamic field, evolving and adapting to meet the challenges of new technologies and it will continue to do so to address both today’s questions and those of tomorrow"

Within the biopharmaceutical industry, Immunocore is a company dedicated to fighting diseases such as cancer with treatment approaches that utilize a patient’s own immune system to attack tumors. Bioinformatics is a cornerstone to this research, providing novel algorithms and comprehensive databases to extend these therapies to target additional biological pathways and identify patients most likely to benefit from treatment.

As our understanding of human disease continues to improve, bioinformatics will play a significant role in the process. Our current concept of ‘big data’ will be considered modest as innovation produces data at orders of magnitude beyond existing standards. Bioinformaticsis a dynamic field, evolving and adapting to meet the challenges of new technologies and it will continue to do so to address both today’s questions and those of tomorrow.