The average cost of launching a new drug is enormous – it can be as high as $2.6 billion. If we consider how many different medications people need to treat diseases (and how many are still in the making) the investment is overwhelming. It is, therefore, hardly surprising that pharmaceutical companies are turning to AI to accelerate drug formulation and expedite clinical trials. The potential of these technologies is remarkable, as we explore below.  

Key Challenges in AI Implementation And How to address Them

EFFECTIVE INTERDISCIPLINARY COLLABORATION 

Ensuring that you have a team of domain experts, who ideally complement each other’s skills is as important in AI implementations as in other business projects. Yet, while pharma experts are highly skilled in drug discovery and development, they might lack AI expertise. 

To bridge this gap, pharma companies can bring in AI specialists and foster an ongoing dialogue between domain experts and data scientists. This encourages the sharing of knowledge and ensures AI is applied in ways that align with pharmaceutical goals. 

Karol Bubała, Scientific Software Developer at Holisticon Connect says: 

“Both worlds – scientists and developers – have something valuable to offer. Engineers know how to create stable, maintainable, and scalable solutions. However, they often need domain-specific expertise”.  “On the other hand, scientists develop solutions that address specific problems, but they typically lack knowledge about software development best practices”, he underlines. “When a large company acquires such software and intends to turn it into a product, it must reconcile these two worlds and find common ground”. 

SCALABILITY OF AI SOLUTIONS

There are a few technical challenges that could hinder your AI project’s move from pilot stages to full-scale deployment. For starters, the more complex your system, the harder it can be to manage data effectively. This can become an issue early on, as training AI models at scale requires vast amounts of clean, consistent, and accessible data. 

To prevent these blockers, it’s important to build automated data pipelines that can handle data cleaning and processing before it’s used by your model. It’s also worth turning to cloud-based infrastructure. These services offer pay-as-you-go schemes, which eliminates the need of making significant upfront investments on hardware. Businesses can only pay for the resources they consume and scale as needed.  

SAFEGUARDING DATA QUALITY AND AVAILABILITY  

According to Statista, in 2024 alone, 147 zettabytes of data were generated. This can be both good and bad news. On the one hand, businesses gain access to a lot of information, especially if they’re willing to exchange data. But the truth is, if most data is fragmented, it’s hard to create a unified dataset, which has a negative impact on data quality. Also, incomplete and inconsistent data can reduce the effectiveness of AI models.  

After the introduction of regulatory acts like GDPR and HIPAA to protect highly-sensitive data, organisations have to pay even more attention to data handling and sharing to avoid any data breaches. 

Is there anything that companies can do to make the most of available data, even when it’s fragmented? Yes, they can adopt data integration platforms that consolidate data from multiple sources into a centralised repository. To ensure accuracy and consistency of datasets, businesses can consider implementing data cleaning pipelines.  

In terms of AI in pharma, data governance frameworks play an important role – they’re vital in guaranteeing data lineage, quality, and compliance. 

Benefits of using AI in pharma along with real-life examples

FASTER, SMARTER, BETTER: AI IN DRUG DISCOVERY

It can take up to 15 years and cost between $1 to 2 billion to discover a single drug using the traditional drug discovery process.

This is because of growing attrition rates and the length of clinical trials. And the worst is, that despite all the time and effort, almost 90% of potential drug candidates fail.   

AI and machine learning (ML) have transformed drug discovery, making tasks like identifying chemical compounds, predicting drug toxicity, monitoring efficacy, and assessing physicochemical properties more achievable.  

The potential of AI in drug discovery is huge. It can analyse data from large populations to spot trends and patterns which can help predict how effective potential drug candidates can be for specific populations. This can assist in tailoring treatments to the needs of individual patients.  

AstraZeneca and Recursion Pharmaceuticals are two examples of companies that use AI in drug discovery. Let’s take a closer look at each.  

Examples:

AstraZeneca

Their pathologists dedicate a lot of hours to analysing hundreds of tissue samples from their research studies. They examine them for diseases and to identify biomarkers that could signal which patients are most likely to benefit from their medications. 

To make this process more effective, AstraZeneca trains AI systems in analysing samples accurately and more effortlessly, potentially reducing the analysis time by over 30%. The company said that one of their AI systems uses an approach inspired by how some self-driving cars understand their surroundings. The AI was trained to assess tumour and immune cells for a biomarker called PD-L1, which could help guide immunotherapy treatment decisions for bladder cancer. 

As of today, AstraZeneca uses AI across 70% of their small molecule chemistry projects to pick the best molecules to make in the shortest time possible.

Recursion Pharmaceuticals 

To speed up the discovery of treatments of rare, untreated diseases, Recursion Pharmaceuticals is working on an AI-enabled human biology map. It combines automated biology, chemistry, and cloud computing tools to uncover new therapeutic candidates, potentially reducing the time needed to discover and develop a new medicine by up to tenfold. 

The company is already seeing results. They have created hundreds of rare disease models and generated a shortlist of drug candidates across multiple diseases (including cerebral cavernous malformation) – all of this in less than 2 years.  

FROM PIXELS TO PRECISION: AI IN MEDICAL IMAGING

Medical imaging analysis is, arguably, one of the most publicised AI healthcare use cases in recent years. Studies conducted during the COVID-19 pandemic showed that AI models were able to recognize lung infection areas at an accuracy rate of 80.19% to 98.91%. Radiologists are also using these systems to improve proper diagnoses for “subtle findings” – from a AUC (Area Under the Curve) rate of 0.57 without AI assistance to 0.92 with AI support​. 

Pharma brands Bayer and Johnson & Johnson are among the pioneers in these AI in pharma technologies. 

Examples:

Calantic by Bayer

Calantic supports radiologists in CT scan analyses and diagnostics. The model comes with five modules, each specialised on individual body parts, i.e., thoracic, neuro, prostate, breast, and cardiac modules. What makes Bayer’s solution particularly noteworthy is that it can also automatically prioritise suspected findings. As a result, radiologists and disease researchers can focus their work on time-critical cases.  

Johnson & Johnson 

In October 2024, the company announced that they were working on an AI-powered biomarker test that could be used for biopsy analyses. The model is trained to recognize potential bladder cancer cases with FGFR alterations.  

This in-the-making solution comes as one of other groundbreaking Johnson & Johnson technologies in the pharma and medical field, after the 3D heart-mapping CARTO 3 system and Polyphonic used for pre-surgery analyses. 

The goal for J&J’s bladder cancer AI test is to facilitate personalised treatments and thus boost chances of a patient’s full recovery. 

FROM LAB TO MARKET: HOW AI SPEEDS UP DRUG DEVELOPMENT

AI offers significant benefits not only in discovering promising compounds and formulations, but also assessing their potential efficacy and accelerating their time-to-market.  

Among others, drug development AI technologies can be used to draw the patient profile for clinical trial use. This is possible thanks to large scale patient data analyses and running predictive modelling to forecast possible clinical trial outcomes.  

Throughout the trial process, these models can help refine dosage strategies based on individual and group patient responses, as well as reducing side effects through pharmacokinetics and pharmacodynamics analyses.  

Once a drug has been cleared for the market, AI can also be used to predict drug demand and any variables that could impact drug quality. Many of these models also run post-market, real-world data analyses to spot previously-unknown adverse effects and ensure patient safety. 

In this case, AI helps pharma brands accelerate bringing newly-formulated treatments to the market. Companies that form strategic partnerships often lead the way in bringing groundbreaking discoveries to patients – and Novartis is one brand that certainly took note. 

Example:

Novartis

Novartis entered a multi-target collaboration with Generate:Biomedicines, who developed a genAI platform for drug molecule formulations. The objective of this joint initiative is to discover and “develop protein therapeutics across multiple disease areas”. 

The pharma giant has announced that they’ll combine the Generate:Biomedicines ML platform to validate treatment hypotheses and experiments across biologics development, clinical development, and target biology.   

GROUNDBREAKING GENE/MUTATION DISCOVERIES THROUGH GENOMIC AND PROTEOMIC ANALYSIS

AI simplifies proteomic and genomics data analysis by quickly (and accurately) processing vast amounts of complex biological data. In genomics, AI tools help with tasks like sequence alignment and annotation, which allows scientists to identify genes, mutations, and regulatory elements in DNA.  

These insights are key to understanding genetic variations linked to diseases, paving the way for more precise treatments. 

In proteomics, AI steps in to analyse protein structures and interactions, using methods like deep learning to predict folding patterns and functional sites. This helps identify potential drug targets and sheds light on protein-protein interactions and pathways, revealing the mechanisms behind cellular processes. 

Example:

Sanofi 

Sanofi entered into a collaboration with SandboxAQ, developers of quantitative AI models, to extend their genomic and proteomic analysis capabilities. As a result of the partnership, the pharma brand can identify new disease biomarkers during clinical development and ideate personalised therapies. 

As per SandboxAQ’s press announcement, “quantitative AI models aim to understand human biology, assisting in the identification of new biomarkers and aiding scientists’ ability to demonstrate the mechanism of action, efficacy, and safety of investigational medicines and targets in clinical development.” 

SMARTER MANUFACTURING PROCESSES: HOW AI OPTIMISE PHARMA PRODUCTION

Another benefit of using AI in pharma is the optimisation of manufacturing processes. AI enables manufacturers to reduce downtime and improve process efficiency by performing predictive maintenance. Even the slightest delay in the production or distribution of medicine can have severe consequences for both patients who need their drugs and for the supply chain.  

Unfortunately, as many as 83% of pharma plants still use preventative maintenance approaches, like over – maintenance, which is simply inefficient. Through the use of AI and data science in pharma, operational teams can quickly check machine health, process efficiency, and other data in real time. Proactive maintenance can aid in minimising machine downtime and safeguarding operations, all the while maintaining high production quality. What does it look like in practice? Let’s see. 

Example:

GlaxoSmithKline (GSK) 

To improve flexibility and production speed, in 2019 GSK invested $120 million to create a state-of-the-art manufacturing hub that uses AI and ML. Data, analytics, and insights from different site areas are consolidated in an advanced digital control room, equipped with touch screens on the walls. The site personnel can monitor operations and gain a deeper understanding of process efficiency. This lets the site reduce deviations and boost product yields, ultimately accelerating the delivery of treatments to those in need. By using AI in vaccine production facilities, GSK was able to achieve a 30% increase in yield.  

Through the use of AI and ML models, the manufacturing team knows when the equipment will need to be replaced and what changes will have to be applied in the future to minimise the number of rejected products.  

AI IN Pharma CAN TRUELY SAVE LIVES

Arguably, few industries can benefit from AI as strongly as pharma and the broader healthcare sector. These technologies not only have the potential to cut costs and accelerate drug development but, most importantly, they can save lives and make a lasting impact on the world.

That said, the biggest discoveries take place when scientists and technology experts join forces – as we saw in the case of Novartis and Sanofi, who turned to specialised genAI and ML platforms to bring new technologies and treatments to the market.  

Passion and Execution

About Holisticon Connect

At Holisticon Connect, our core values of Passion and Execution drive us toward a Promising Future. We are a hands-on tech company that places people at the centre of everything we do. Specializing in Custom Software Development, Cloud and Operations, Bespoke Data Visualisations, Engineering & Embedded services, we build trust through our promise to deliver and a no-drama approach. We are committed to delivering reliable and effective solutions, ensuring our clients can count on us to meet their needs with integrity and excellence. 

Learn more