Charting New Horizons: The Expanding Role of AI in Healthcare and Protein Engineering

In the current day and age, big data and machine learning play a crucial role in several sectors, including entertainment, commerce, and healthcare. Big data refers to the large volumes of data that are collected, analyzed, and processed to extract insights and information. Machine learning, a subset of Artificial Intelligence (AI), involves the use of algorithms and statistical models to enable computer systems to learn from data and improve performance without being explicitly programmed [1-3]. AI also involves the development of computer systems that possess human-like intelligence, like problem-solving, reasoning, learning, perception, and decision-making.

The vast domain of healthcare is increasingly recognizing the importance of incorporating such AI-powered tools that encompass Machine Learning (ML) and big data into the nextgeneration healthcare technology holding the potential to significantly improve healthcare operations and delivery. They could greatly benefit from the potential improvements AI has to offer [3-5]. For instance, AI applications are estimated to reduce annual healthcare costs in the United States by approximately USD 150 billion by 2026. These cost reductions primarily stem from transitioning the healthcare model from a reactive approach focused on treating diseases to a proactive approach centered around health management, while also leveraging AI-based technology, continuous monitoring, coaching, early diagnosis, tailored treatments and efficient followups can be prioritized, enabling individuals to maintain better overall health.

In recent years, there has been a notable surge in interest surrounding generative AI, a technology capable of generating new data based on existing data, and its potential applications across various fields. The broad potential of generative AI allows us to produce a diverse range of outputs [6]. However, it is crucial to exercise caution and conduct a thorough analysis, considering the significant opportunities and implications associated with AI. This careful evaluation holds particular importance in industries such as pharmaceuticals, known for their slow pace of change and stringent regulatory environment. However, AI-enabled technology is also actively addressing crucial issues in drug discovery and development, leveraging advancements, growth, and the abundance of available data to generate meaningful insights. Consequently, there is a need for comprehensive research articles that examine the contributions of AI, machine learning, and big data to drug discovery and development [7]. By incorporating the latest AI works into practical pharmaceutical industry applications, the future holds promising possibilities.

The world of artificial intelligence is a complex and dynamic one, with the market reaching an astounding value of 142.3 billion U.S. dollars as of 2023. This impressive figure is a testament to the continued influx of investments in the field, driving its rapid expansion from billions to trillions of U.S. dollars in the near future. The global surge in investments in AI startups between 2020 and 2022, which almost doubled previous investment figures, highlights the growing interest in this innovative and transformative industry. It is fascinating to note that, although the fastest growing market is South-East Asia and Pacific, a significant portion of these investments came from private capital provided by U.S. companies, indicating the country’s leadership in the AI landscape.

For instance, in January 2023, “Profluent” emerged with $9 million in seed funding to advance its application of Machine Learning (ML) for engineering functional and viable proteins. Last year, “Generate Biomedicines” secured a $50 million drug development deal with Amgen, potentially earning over $1.9 billion in total. Similarly, “Arzeda” received $33 million in series B funding to support its protein design programs [9-11]. The space is becoming increasingly competitive, with newcomers like “Cradle”, a computational company, raising $5.5 million in seed investment after exiting stealth mode, and “Monod Bio”, which launched with $25 million in seed funding last August.

AI holds great potential in the pharmaceutical industry, offering opportunities to enhance cost-effectiveness and efficiency. Extensive research has shown that dynamic learning approaches can create highly accurate AI models using significantly less data compared to traditional methods and data subsampling techniques. While the precise reasons behind this increased productivity are not yet fully understood, reduced redundancy, bias, and the ability to gather more substantial information for decision-making appear to be key factors contributing to improved performance. As a result, screening costs have the potential to be reduced by up to 90%, disregarding any expected overhead associated with implementing dynamic learning efforts.

Machine learning techniques, particularly deep learning, have proven effective in managing complex analyses of large, diverse, and high-dimensional datasets without requiring manual input. When combined with human expertise and experience, machine learning becomes a powerful tool for efficiently handling vast amounts of data. This capability for information mining has brought new significance to computer-assisted drug design, where the integration of multiple clinical factors into a cohesive framework surpasses fragmentary information, accelerating the drug discovery process.

With the continued accumulation of clinical data and advancements in AI algorithms, AI technology is poised to revolutionize various aspects of medicine. The synergistic progress of automation and the integration of technologies is expected to drive advances in medication development by enabling improved analysis of large and complex datasets. As the world of AI continues to evolve and expand, the possibilities and opportunities are endless, making it an exciting and promising field to watch. The number of applications keeps on rising and there is a growing interest and potential in applying ML in medicine and protein engineering.

The field of medical diagnostics involves a comprehensive evaluation of medical conditions, taking into account symptoms, medical history, and test results. Its primary idea is to directly identify the underpinning cause of a medical problem, allowing for precise judgments and effective treatment strategies. A wide array of individual tests, including imaging tests such as X-ray, MRI and CT reviews, as well as blood tests, give precious perceptivity that props healthcare providers in determining the most suitable approach for treatment. Medical diagnostics also plays a vital part in covering the progress of conditions, assessing the effectiveness of treatments, and detecting implicit health issues at an early stage.

The recent arrival of AI has brought forth an implicit revolution in the realm of medical diagnostics, promising advancements in speed and effectiveness. AI algorithms are able to validate and process different types of patient data, encompassing medical imaging, as well as ECG, EEG, EMG, and EHR, and vital signs including body temperature, palpitation rate, respiration rate, and blood pressure. Likewise, demographic information, medical history, and laboratory test results can also be integrated within the ML. This admixture of different data sources provides an intelligent result that improves individual decision-making through a multifaceted approach. By considering multiple sources of information, healthcare providers can gain a comprehensive understanding of a patient’s health, significantly reducing the liability of misdiagnosis and enhancing personalization. The application of multimodal data allows for nonstop monitoring of condition progression and facilitates more effective operation and treatment of habitual conditions. A fresh benefit of AI, particularly XAI, is its capability to detect implicit health issues at an earlier stage, therefore easing or even precluding potentially life-changing situations.

AI is being used as a tool for diagnosis allowing 24/7 availability, a very low percentage of errors, an ability to provide real-time insights, and performing fast analyses. For example, quantum computers possess remarkable processing power allowing for realtime analysis of vast amounts of medical data to ensure accurate diagnoses. Quantum optimization algorithms are used to optimize decision-making in medical diagnostics by considering medical history and other pertinent factors. Another concept known as General AI (GAI) is being employed by various projects and companies such as Open AIs DeepQA, IBMs Watson, and Google’s DeepMind. GAI aims to enhance the accuracy, speed, and efficiency of diagnoses while providing valuable insights and support to healthcare providers, for example by generating personalized treatment plans using predictive analytics.

The future of AI-driven medical diagnostics is expected to experience ongoing growth and advancement. Cutting-edge AI technologies, including Quantum AI (QAI) are being introduced to expedite training and provide quick diagnostic models. The rapid advances in Artificial Intelligence (AI) have sparked the emergence of sophisticated Large Language Models (LLMs), exemplified by the likes of GPT-4 and Bard. These LLMs have garnered considerable attention within the healthcare domain due to their versatile applications. Their potential implementation spans a broad spectrum, ranging from facilitating clinical documentation and streamlining insurance pre-authorization processes to summarizing extensive research papers. Furthermore, they can even function as chatbots, addressing patients’ questions and concerns regarding their personal health data. While these LLMs hold tremendous transformative potential, they necessitate a cautious approach. Their training methodologies differ from those employed by regulated AI-based medical technologies, particularly when within the critical context of caring for patients.

In summary, AI is revolutionizing the clinical field by being so versatile and offering support in multiple applications including disease diagnosis, risk assessment, treatment planning, and drug discovery resulting in improved patient outcomes and a more efficient healthcare system. AI is also helping the drug discovery process by automating and optimizing target identification, hit identification, lead optimization, chemical synthesis prediction, and drug repositioning. Another study Jalil Villalobos-Alva et al. [12] showcased a systematic review and biochemical meta-analysis focused on the tremendous innovation of binomial studies in both AI applications and guidance in drug design and protein engineering. The contribution of this study was threefold:

First, the creation of a cross-functional team where computer scientists, biomedical professionals and philosophers created a common language and together identified relevant literature in the various fields of AI and PS and built a bridge between the two areas that can serve as a framework for further research in both fields.

Second, the study emphasized the importance of a better understanding of ML model training and validation methods and their results by combining databases from multiple areas of knowledge (e.g., biological experiments, in-silico simulations, ML, direct evolutionary approach) that allowed us to classify, layer and layer to contribute to the emerging field of protein science. Third, the study showed that binomial AI-PS, a progressive research program, still has several challenges to overcome, such as the development of a comprehensive machine learning reference, and experimental validation of the 3D modeling structure. In laboratories, classification, etc. to control the vulnerability of neural networks, developing a tool to design new biocatalysts not found in nature using reverse engineering, man-made metabolic pathways, designing a new antibody molecular factory, new proteostasis systems, understanding the mechanisms of protein folding and protein aggregation [13].

Finally, the study proposed that the recent great success of AIPS- Alphafold may lead to a paradigm shift in research, encouraging a new generation of researchers to use it. In any case, it is clear that a cross-functional group of researchers from multiple fields of knowledge must work in concert to share ideas, methods, and challenges to develop explicit computational tools and paradigms, to fully explore and close the gap in Binomial AI-PS, a promising research area.

Enzymes are used to synthesize drugs on a large scale, often requiring them to function as efficient catalysts under non-optimal conditions. Directed evolution of proteins involves multiple rounds of experiments resembling natural selection, where mutated sequences are created and evaluated for desired properties. However, identifying favorable mutations within the vast protein sequence space remains challenging, despite advancements in the field. Machine learning models have the potential to learn from existing sequences and protein properties. They have gained traction as a tool to aid protein redesign by virtually screening numerous novel sequences. They can also be trained to make predictions on new mutations to be synthesized or properties of newly synthesized proteins.

The study benchmarks the performance of prediction models (for single amino acid descriptors and one structure-based threedimensional descriptor) constructed using various machinelearning methods and protein descriptors. The evaluation utilizes diverse datasets, including public and proprietary ones, and employs different evaluation metrics. The results indicate that Convolutional Neural Network (CNN) models, employing amino acid property descriptors, are highly applicable to pharmaceutical industry challenges in protein redesign.

Protein optimization through computational approaches, particularly machine learning models, has accumulated increasing attention for guiding the design of experiments and synthesizing promising protein sequences. Machine learning-based protein optimization relies on appropriate descriptors that effectively capture the information present in protein sequences. Various protein sequence descriptors have been successfully employed in sequence-based protein classification and ligand docking tasks. These descriptors include: -
A. Amino acid composition (dipeptide, tripeptide composition)
B. Predicted secondary structure.
C. Predicted solvent accessibility.
D. Histograms of torsion angles density and amino acid distances density
E. k-Spaced Amino Acid Pairs and Conjoint Triad
F. 3D grid protein-ligand structures

AI-based protein engineering has been pushed to the next level with AI Proteins, a biotechnology firm that uses artificial intelligence to design synthetic proteins de novo. They secured a $18.2 million seed round that was oversubscribed and co-led by Cobro Ventures and Light chain Capital. With a cutting-edge method for creating entirely novel proteins, AI Proteins is rethinking protein treatments. De novo proteins, that are created at AI Proteins, use AI-based design and a high-throughput drug discovery platform to be optimized for certain therapeutic uses. The AI Proteins platform makes it possible to create low-cost, enduring, highly specific proteins that may be tailored for oral administration.

Protein engineering does also pose a significant concern regarding the enhancement of known protein toxins or the creation of new ‘designer’ toxins for hostile purposes. These toxins have been sought after for warfare, assassination, and other destructive intentions. However, protein toxins also have beneficial applications due to their ability to specifically recognize and bind to molecular targets. A worrisome area within protein engineering is the development of hybrid or fusion toxins for advanced therapeutics. By employing protein engineering techniques, the effectiveness of fusion toxins in eliminating target cells can be significantly increased.

For instance, a study demonstrated a 17-fold increase in cytotoxicity using these techniques. Although most protein toxin research is conducted for peaceful purposes and shared in scientific publications, there is a risk of misuse of biological weapons. Numerous documented cases involve the acquisition of protein toxins for hostile activities. More than a decade ago, The World Armaments and Disarmament Yearbook highlighted the military interest in the ease of producing novel engineered bacterial toxins and similar substances through protein engineering. Moreover, protein engineering holds the speculative potential to convert harmless proteins into toxins by manipulating their interactions with different substrates, proteins, or DNA sequences. Researchers in protein toxins must be aware of their legal and ethical responsibilities under the Biological and Toxin Weapons Convention (BTWC) and the Chemical Weapons Convention (CWC). They should educate themselves and the next generation of researchers about these obligations and report suspicious activities that could enhance toxin lethality or military applicability. Countries engaged in protein toxin engineering should establish national biosecurity boards to oversee proposed experiments, similar to regulations governing research on human and animal pathogens.

Although scientific interests and global investments are constantly increasing, the use and development of AI are under strict scrutiny. Only some ML algorithms have been clinically approved because essential data privacy and security considerations, along with ethical and legal concerns need to be considered. The regulation of responsible design, development, and use of AI in healthcare is still in its early stages due to the rapid evolution of the field. Therefore, the gap between what is technically possible and what is allowed by privacy legislation is growing. Inappropriately implementing new technologies can pose substantial risks. A pertinent example is the initial excitement surrounding IBM’s Watson Health, which was anticipated to have the ability to detect complex cancers but ultimately fell short of meeting expectations.e mechanisms of protein folding and protein aggregation [13].

The development and implementation of AI in medical diagnostics encounter technical, regulatory, and ethical challenges. One challenge is the availability of high-quality labelled data since medical data can be fragmented, incomplete, unlabeled, or inaccessible altogether. If AI algorithms are trained on nonrepresentative data. There is a risk of bias leading to incorrect or unfair diagnoses. Ethical concerns arise about using AI with private and sensitive datasets that encompass issues like data privacy, algorithmic transparency, and accountability. While solutions like federated learning have been proposed as possible remedies for these challenges, further investigation is needed to ensure their effectiveness in the field of medical research. Lastly, interoperability standards and protocols are crucial for facilitating effective collaboration between different entities that develop AIbased medical diagnostic tools.

In the Protein Engineering field, progress is constantly being made. For example, ML-based protein-ligand interactions keep on using a larger number of physicochemical properties, and the latest versions of ML algorithms are used to always achieve more accurate predictions. Both the size and the quality of the datasets are important because incomplete and partial information would cause poor, inaccurate, and biased predictions. Improvement in the quality of AI predictions is dependent on a constant feed-forward loop between predictions and experimental data.

To conclude, the main challenges for the AI industry are A. Its versatility, i.e., the technological development to apply AI to as many fields as possible.
B. The current technological limitations.
C. Ethics, i.e., some developments and applications can put people’s privacy at risk or create problems regarding eugenics, calling for some guidance and policymaking as well as education of the general public to generate and nurture trust.

1. Al-Antari MA (2023) Artificial intelligence for medical diagnostics-existing and future AI technology. Diagnostics (Basel) 13(4): 688
2. Arowora, Kayode A (2022) Application of artificial intelligence in biochemistry and biochemical sciences: A Review Asian Res J Curr Sci 4(1): 302-312.
3. Bohr A, Memarzadeh K (2020) The rise of artificial intelligence in healthcare applications. Artificial Intelligence in Healthcare 25-60.
4. Chen Z, Yu Y, Gao Y, Zhu Z (2023) Rational design strategies for nanozymes. ACS Nano 17(14): 13062-13080.
5. Dhakal A, Kay C, Tanner JJ, Cheng J (2022) Artificial intelligence in the prediction of protein-ligand interactions: recent advances and future directions. Brief Bioinform 23(1): bbab476.
6. Eisenstein M (2023) AI-enhanced protein design makes proteins that have never existed. Nat Biotechnol 41(3): 303-305.
7. Engqvist M, Rabe KS (2019) Applications of protein engineering and directed evolution in plant research. Plant Physiology 179(3): 907-917.
8. Meskó B, Topol EJ (2023) The imperative for regulatory oversight of large language models (or generative AI) in healthcare. NPJ Digit Med. 6(1): 120.
9. Patel V, Shah M (20232) Artificial intelligence and machine learning in drug discovery and development. Intel Med 2(3): 134-140.
10. Stafie CS, Sufaru IG, Ghiciuc CM, Stafie II, Sufaru EC (2023) Exploring the intersection of artificial intelligence and clinical healthcare: A multidisciplinary review. Diagnostics (Basel) 13(12): 1995.
11. Tucker JB, Hooper C (2006) Protein engineering: security implications. The increasing ability to manipulate protein toxins for hostile purposes has prompted calls for regulation. EMBO Rep Spec No(Spec No): S14-7.
12. Villalobos AJ, Ochoa TL (2022) Protein science meets artificial intelligence: A Systematic review and a biochemical meta-analysis of an inter-field. Front Bioeng Biotecnol 10: 788300.
13. Xu Y, Verma D (2020) Deep dive into machine learning models for protein engineering. J Chen Inf Model 60(6): 2773-2790.