AI-powered drug discovery is revolutionizing the pharmaceutical industry by dramatically accelerating the process of identifying potential drug compounds, understanding their interactions, and predicting their efficacy and safety. Traditionally, drug discovery has been a lengthy, expensive, and high-risk endeavor, often taking over a decade and billions of dollars with a high failure rate. AI promises to mitigate these challenges by leveraging vast datasets and advanced computational models.

In India, an EY report (Feb 2025) indicated that 50% of Indian Pharma companies are exploring or investing in AI-driven solutions, with 25% already having Generative AI applications in production. This signifies a strong push towards leveraging AI to move beyond generics and drive novel drug development.

Here’s a breakdown of the industrial application of AI in drug discovery:

1. Target Identification and Validation

• Traditional Challenge: Identifying the specific biological molecules (targets like proteins, enzymes, or genes) whose modulation can treat a disease is the first, often arduous, step. Many potential targets are explored, leading to significant time and resource investment.
• AI Application:
  • Genomic and Proteomic Analysis: AI analyzes vast datasets from genomics, proteomics, transcriptomics, and metabolomics. It uses machine learning algorithms to identify disease-associated genes, proteins, and biological pathways that are most likely to be effective drug targets.
  • Literature Mining and Knowledge Graphs: Natural Language Processing (NLP) and graph neural networks can extract insights from millions of scientific papers, patents, and clinical trial reports to build knowledge graphs that connect diseases, genes, proteins, and potential compounds, highlighting promising new targets.
  • Predicting Protein Structures: AI systems like Google DeepMind’s AlphaFold (Nobel Prize in Chemistry 2024 for computational protein design and structure prediction) accurately predict the 3D structures of proteins from their amino acid sequences. Understanding these structures is crucial for designing drugs that precisely interact with them.
• Industrial Impact: Dramatically reduces the time and cost associated with identifying and validating the most promising disease targets, allowing companies to focus their resources more efficiently.

2. Lead Discovery and Hit Identification (Virtual Screening)

• Traditional Challenge: Screening millions of compounds in “wet labs” to find those that show any initial activity (hits) against a chosen target is labor-intensive, time-consuming, and expensive.
• AI Application:
  • High-Throughput Virtual Screening: AI algorithms (e.g., deep learning models, molecular docking simulations) can rapidly screen vast virtual libraries of billions of chemical compounds against a target protein structure. They predict how strongly and precisely a compound might bind to the target. This “in-silico” (computer-simulated) screening drastically reduces the need for physical experimentation.
  • Generative AI for De Novo Design: Generative AI models (like Generative Adversarial Networks – GANs or Variational Autoencoders – VAEs) can design entirely new chemical compounds from scratch. Given specific desired properties (e.g., binding affinity to a target, low toxicity, solubility), these models can generate novel molecular structures that have never existed before, tailored to fit the target’s binding site.
• Industrial Impact: Shortens the hit identification phase from years to months, significantly cuts laboratory costs, and expands the chemical space explored beyond what’s physically feasible.

3. Lead Optimization and Property Prediction

• Traditional Challenge: Initial “hits” often have undesirable properties (e.g., poor solubility, high toxicity, low bioavailability). Medicinal chemists spend significant time modifying these compounds to improve their drug-like qualities. This is an iterative and often unpredictable process.
• AI Application:
  • Predicting ADMET Properties: AI models predict critical ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) properties of potential drug candidates based on their chemical structure. This helps filter out compounds likely to fail in later stages due to poor pharmacokinetics or safety concerns.
  • Structure-Activity Relationship (SAR) Modeling: AI builds complex models to understand how small changes in a chemical structure affect its biological activity. This guides medicinal chemists in optimizing leads for potency, selectivity, and safety.
  • Synthetic Route Prediction: AI can predict feasible chemical synthesis routes for newly designed or optimized compounds, ensuring that the promising molecules can actually be manufactured.
• Industrial Impact: Reduces the number of compounds that need to be synthesized and tested, accelerates the optimization cycle, and increases the likelihood of a compound successfully moving into preclinical development.

4. Drug Repurposing (Repositioning)

• Traditional Challenge: Finding new therapeutic uses for existing, approved drugs is difficult due to the sheer volume of drugs and diseases.
• AI Application:
  • Analyzing Multi-Omics and Clinical Data: AI analyzes vast datasets, including patient electronic health records (EHRs), genomic data, real-world evidence, and existing drug databases. It identifies hidden connections between drugs, their mechanisms of action, and different disease pathways.
  • Predicting New Indications: AI can predict that a drug approved for one condition might be effective against another. For example, during the COVID-19 pandemic, AI was instrumental in identifying potential antiviral candidates among existing drugs.
• Industrial Impact: Provides a faster, less risky, and more cost-effective path to new therapies, as the repurposed drugs have already undergone extensive safety testing.

5. Preclinical and Clinical Trial Optimization

• Traditional Challenge: Clinical trials are the most expensive and time-consuming part of drug development, with a high failure rate. Patient recruitment, site selection, and data analysis are complex.
• AI Application:
  • Patient Stratification and Recruitment: AI analyzes patient data (genetics, medical history, lifestyle) to identify ideal patient populations for clinical trials, making recruitment faster and ensuring a higher probability of success.
  • Predicting Trial Outcomes: AI can predict the likelihood of success or potential adverse events in clinical trials by analyzing preclinical data, patient demographics, and historical trial outcomes.
  • Trial Design Optimization: AI helps design more efficient and effective clinical trials by optimizing dosages, endpoints, and study duration.
  • Biomarker Discovery: AI identifies biomarkers that predict drug response or disease progression, leading to personalized medicine approaches.
• Industrial Impact: Reduces the high failure rate of clinical trials, shortens trial timelines, lowers costs, and helps bring safer, more effective drugs to market faster.

Indian Companies and Initiatives in AI-Powered Drug Discovery:

India, with its strong pharmaceutical base (often referred to as the “pharmacy of the world” for generics), is increasingly investing in AI for novel drug development.

• Startups: Several Indian startups are emerging in this space, often focusing on specific therapeutic areas or AI methodologies. While the landscape is evolving rapidly, companies are looking at AI to enhance their R&D capabilities.
• Established Pharma Players: Major Indian pharmaceutical companies like Sun Pharma, Dr. Reddy’s Laboratories, Lupin, Cipla, Biocon, Aurobindo Pharma, and Glenmark Pharmaceuticals are integrating AI into their R&D processes, either through in-house capabilities, partnerships with AI companies (both domestic and international), or by exploring AI for aspects like manufacturing optimization and supply chain management.
  • For instance, the Indian Council of Medical Research (ICMR) has collaborated with AI-powered platforms to accelerate the development of tuberculosis (TB) treatments.
• Academic and Research Institutions: Indian institutes like IITs, IISc, and various universities are actively engaged in research at the intersection of AI, computational chemistry, and biology, contributing to the foundational knowledge for AI-driven drug discovery.
• Government Support: The Indian government’s push for digital transformation and support for AI initiatives (e.g., National Strategy for AI) implicitly encourages AI adoption in crucial sectors like pharmaceuticals.

Conclusion:

AI-powered drug discovery is no longer a distant dream but a tangible industrial application that is reshaping the pharmaceutical landscape. By dramatically improving efficiency, accuracy, and speed across the entire drug discovery pipeline, AI is enabling the identification of novel compounds, accelerating the development of new medicines, and ultimately bringing life-saving treatments to patients faster and more cost-effectively. India’s growing investment and talent pool in both pharma and AI position it well to become a significant player in this transformative field.

What is AI-Powered Drug Discovery – Using AI to identify potential pharmaceutical compounds?

AI-powered drug discovery refers to the application of Artificial Intelligence (AI) technologies, primarily Machine Learning (ML) and Deep Learning (DL), to significantly accelerate, de-risk, and optimize the process of finding and developing new pharmaceutical compounds. It aims to overcome the traditional challenges of drug discovery, which are known to be extremely lengthy (10-15 years), costly (billions of dollars per drug), and characterized by high failure rates.

In essence, AI helps identify potential pharmaceutical compounds by:

1. Analyzing Vast Datasets: AI can process and extract insights from enormous and diverse datasets that are impossible for humans to handle. This includes:
   • Chemical Data: Structures of millions, even billions, of known and theoretical compounds.
   • Biological Data: Genomic sequences, proteomic profiles, gene expression data, disease pathways, protein structures (e.g., predicted by AlphaFold).
   • Clinical Data: Electronic health records (EHRs), patient outcomes, clinical trial results, adverse event reports.
   • Scientific Literature: Millions of research papers, patents, and review articles.
2. Identifying Complex Patterns and Relationships: AI algorithms excel at recognizing subtle, non-obvious patterns and correlations within this data that humans or traditional statistical methods might miss. These patterns can indicate:
   • Which biological targets (e.g., specific proteins or genes) are most relevant to a disease.
   • How different chemical compounds interact with these targets.
   • Which compounds are likely to have desired properties (efficacy, potency) and undesirable ones (toxicity, poor absorption).
3. Predicting Key Properties and Outcomes: Based on the patterns learned from data, AI models can make predictions about potential drug candidates, such as:
   • Binding Affinity: How strongly a compound will bind to a specific disease target.
   • ADMET Properties: Prediction of Absorption, Distribution, Metabolism, Excretion, and Toxicity, which are crucial for a drug’s effectiveness and safety.
   • Efficacy: How well a drug might work against a specific disease.
   • Side Effects: Potential adverse reactions a drug might cause.
   • Synthesis Feasibility: How easy or difficult it might be to chemically synthesize a new compound.
   • Clinical Trial Success: Predicting the likelihood of a drug candidate succeeding in human trials.

How AI Identifies Potential Pharmaceutical Compounds – Key Steps:

AI’s role spans the entire drug discovery pipeline, from early-stage research to optimizing clinical trials:

1. Target Identification and Validation:
   • Traditional: Laborious experimental identification of biological molecules (proteins, genes) whose modulation can treat a disease.
   • AI’s Role: AI analyzes genomic, proteomic, and clinical data to pinpoint the most promising disease-causing targets and validate their relevance, dramatically shortening this initial, critical step. AI can also predict the 3D structures of these targets (e.g., AlphaFold), which is vital for drug design.
2. Hit Identification and Lead Discovery (Virtual Screening):
   • Traditional: Physically screening millions of compounds in “wet labs” to find initial “hits” that show some activity against the target.
   • AI’s Role: AI performs virtual screening. It computationally screens billions of chemical compounds against a target protein’s structure, predicting binding affinity and interactions. This drastically reduces the number of physical compounds that need to be synthesized and tested, saving immense time and cost.
   • Generative AI for De Novo Design: Advanced AI models can even design entirely new, novel chemical compounds from scratch that have never existed before, based on desired properties and interactions with the target.
3. Lead Optimization:
   • Traditional: Modifying initial “hit” compounds to improve their drug-like properties (e.g., potency, selectivity, solubility, reduced toxicity). This is an iterative and often unpredictable process.
   • AI’s Role: AI predicts how small changes to a chemical structure will affect its activity and ADMET properties. This guides medicinal chemists in optimizing compounds more efficiently, reducing the trial-and-error approach and accelerating the optimization cycle.
4. Drug Repurposing (Repositioning):
   • Traditional: Accidental or manual discovery of new uses for existing, approved drugs.
   • AI’s Role: AI analyzes vast datasets, including patient records, genomic data, and existing drug databases, to identify hidden connections between existing drugs and new therapeutic indications. This is a faster and less risky path to new therapies since the drugs are already proven safe.
5. Preclinical and Clinical Trial Optimization:
   • Traditional: The most expensive and time-consuming phase, often plagued by high failure rates due to issues like patient recruitment, unexpected toxicity, or lack of efficacy in humans.
   • AI’s Role:
     • Patient Stratification: AI analyzes patient data (genetics, demographics, medical history) to identify ideal patient populations for clinical trials, leading to more targeted and efficient recruitment.
     • Predictive Toxicology and Efficacy: AI can predict the likelihood of adverse events or drug efficacy in humans based on preclinical data, helping to de-risk compounds before they enter expensive human trials.
     • Trial Design Optimization: AI helps optimize trial protocols, dosages, and even identify optimal clinical sites for higher success rates.

By integrating these AI capabilities, drug discovery shifts from a largely empirical, trial-and-error process to a more data-driven, predictive, and intelligent approach. This promises to bring safer, more effective, and more affordable medicines to patients faster than ever before.

Who is require AI-Powered Drug Discovery – Using AI to identify potential pharmaceutical compounds?

Courtesy: BBC StoryWorks

AI-powered drug discovery is not a niche requirement; it’s rapidly becoming a fundamental necessity for any organization aiming to innovate and remain competitive in the pharmaceutical and biotechnology sectors. The traditional drug discovery model is too slow, too expensive, and too prone to failure for the demands of modern medicine.

Here’s a breakdown of who specifically requires AI-powered drug discovery:

1. Large Pharmaceutical Companies (Big Pharma)

• Why: These are the traditional giants of drug development, with massive R&D budgets but also enormous pressure to bring new, blockbuster drugs to market. They face patent cliffs, increasing R&D costs, and a need to replenish their pipelines with novel therapies.
• Specific Needs:
  • Accelerating Pipeline: AI helps them rapidly identify new targets, screen vast compound libraries, and optimize leads, significantly shortening the time to clinical trials.
  • Reducing R&D Costs: By predicting failures early (e.g., toxicity, poor efficacy), AI prevents expensive late-stage clinical trial failures.
  • Exploring Novel Chemistry: Generative AI allows them to design completely new molecules beyond what traditional methods could conceive.
  • Personalized Medicine: AI helps stratify patient populations for more effective clinical trials and identify biomarkers for targeted therapies.
  • Staying Competitive: Competitors (including agile biotech startups) are adopting AI, forcing big pharma to integrate it to maintain their market position.
• Examples (Global & Indian Context): Companies like Pfizer, AstraZeneca, Roche, Novartis, and many Indian pharma giants like Sun Pharma, Dr. Reddy’s Laboratories, Lupin, Cipla, Biocon, and Aurobindo Pharma are heavily investing in or exploring AI-driven solutions. EY reports indicate 50% of Indian pharma companies are exploring or investing in AI, with 25% already having GenAI in production.

2. Biotechnology Companies (Biotech Startups and Established Firms)

• Why: Often more agile and innovation-focused than large pharma, biotechs are driven by groundbreaking science. Many emerging biotechs are founded specifically on AI/ML platforms designed for drug discovery.
• Specific Needs:
  • Rapid Innovation: AI is their core competitive advantage, allowing them to rapidly iterate through drug design cycles.
  • Targeted Therapies: Many biotechs focus on niche diseases or specific biological pathways, where AI can precisely identify targets and design highly selective compounds.
  • Capital Efficiency: AI helps de-risk early-stage programs, making them more attractive to investors and extending their runway.
  • Drug Repurposing: AI can efficiently find new uses for existing drugs, a faster and less capital-intensive path to market.
• Examples (Global): Companies like Insilico Medicine (which claims to have put an AI-designed drug into clinical trials), BenevolentAI, Recursion Pharmaceuticals, Atomwise, Exscientia, Schrödinger, and Deep Genomics are pioneers in this space.
• Examples (Indian Context): Emerging Indian innovators like Cellix Bio, MoleculeAI, and SilicoScientia are making strides, attracting investment and reshaping the industry with AI.

3. Contract Research Organizations (CROs) and Contract Development and Manufacturing Organizations (CDMOs)

• Why: CROs conduct research and clinical trials for pharma/biotech companies, while CDMOs handle drug manufacturing. AI enhances their efficiency and service offerings.
• Specific Needs:
  • Optimized Clinical Trials: AI helps CROs with patient recruitment, site selection, data analysis, and predicting trial outcomes for their clients, leading to faster and more successful trials.
  • Improved Efficiency in Preclinical Services: AI-driven virtual screening and lead optimization tools enhance the quality and speed of services offered to clients.
  • Predictive Manufacturing: CDMOs can use AI to optimize manufacturing processes, predict quality issues, and ensure compliance.
• Examples: Global CROs like IQVIA and Syneos Health are integrating AI. In India, numerous CROs and CDMOs are increasingly adopting AI to enhance their capabilities and attract international clients.

4. Academic and Research Institutions

• Why: Universities and research labs are at the forefront of fundamental biological and chemical discoveries. They generate massive amounts of data that AI can leverage.
• Specific Needs:
  • Unlocking New Knowledge: AI helps researchers sift through complex biological data (genomics, proteomics) to identify novel disease mechanisms and drug targets.
  • Accelerating Basic Research: AI can formulate hypotheses, predict experimental outcomes, and guide laboratory experiments, speeding up the pace of discovery.
  • Training Future Talent: These institutions are crucial for educating the next generation of scientists and data specialists who can apply AI in drug discovery.
• Examples (Indian Context): Institutions like IITs, IISc, NIPER (National Institute of Pharmaceutical Education and Research), and AIIMS are increasingly engaged in AI-driven drug discovery research and collaborations. AIIMS, for instance, has announced significant investment to leverage AI technology across healthcare.

5. Government Bodies and Funding Agencies

• Why: While not directly discovering drugs, these entities play a crucial role in shaping the ecosystem and promoting innovation.
• Specific Needs:
  • Accelerating Public Health Solutions: Governments recognize AI’s potential to quickly develop treatments for pandemics (like COVID-19) or neglected diseases.
  • Funding Innovation: They provide grants and incentives for AI research in drug discovery.
  • Regulatory Frameworks: They are developing guidelines (e.g., India’s DPDP Act, MeitY’s AI Governance Guidelines) to ensure the ethical and responsible use of AI in healthcare and drug development.
• Examples (Indian Context): The Indian Council of Medical Research (ICMR) actively promotes AI use in medical research. The Ministry of Electronics and Information Technology (MeitY) is shaping AI policy, and regulatory bodies are adapting to AI-driven research.

In essence, anyone involved in the pursuit of new medicines, from the largest global corporations to agile startups and cutting-edge academic labs, requires AI-powered drug discovery to stay competitive, efficient, and ultimately, to deliver life-saving treatments to patients faster and more effectively. In India, this is especially true as the nation aims to move beyond its generics stronghold into novel drug innovation.

When is require AI-Powered Drug Discovery – Using AI to identify potential pharmaceutical compounds?

AI-powered drug discovery is not a requirement for “when” it will be needed in the future; it’s a present-day necessity and has been for several years now. Its adoption is rapidly accelerating, and organizations that do not integrate AI into their R&D processes are at a significant disadvantage.

Here’s why the “when” is now, and why its urgency is only increasing:

1. To Overcome the Limitations of Traditional Drug Discovery (Ongoing Imperative):

• Prohibitive Costs: Bringing a new drug to market traditionally costs upwards of $2 billion USD and can reach $6 billion USD for complex areas like oncology. AI helps reduce these costs by accelerating early-stage research and predicting failures earlier.
• Excessive Timelines: The journey from initial discovery to market approval typically takes 10-15 years. AI promises to significantly compress these timelines, potentially cutting them by up to 70% in early phases, reducing years to months. For instance, some AI-designed drugs have moved from target identification to clinical trials in as little as 12-18 months, compared to the traditional 2.5-4 years.
• High Failure Rates: Approximately 90% of drug candidates entering clinical trials fail to reach the market, often due to a lack of efficacy or unforeseen toxicity. AI improves the probability of success by identifying more promising candidates and predicting issues earlier.
• Limited Chemical Space Exploration: Traditional methods struggle to explore the vast “chemical space” (all possible molecules). AI, especially generative AI, can design novel compounds that humans might not conceive, vastly expanding the search for new drugs.

2. To Meet Growing Global Health Challenges (Immediate and Future Needs):

• Emergence of New Diseases: The rapid onset of pandemics (like COVID-19) highlighted the urgent need for accelerated drug development. AI proved invaluable in repurposing existing drugs and identifying potential new therapies quickly.
• Antibiotic Resistance: As antibiotic resistance grows, AI is crucial for discovering entirely new classes of antimicrobial compounds.
• Rare Diseases: For diseases affecting small patient populations, traditional R&D models often aren’t financially viable. AI can make drug discovery more cost-effective for these neglected indications.
• Aging Populations and Chronic Diseases: The increasing prevalence of chronic diseases and an aging global population demand a faster pipeline of effective treatments.

3. To Stay Competitive in a Rapidly Evolving Industry (Current Market Imperative):

• Early Adopters Gaining an Edge: Companies that embraced AI early are already seeing promising results, with several AI-designed drugs now in clinical trials (e.g., Insilico Medicine’s drug for IPF in Phase II trials). This demonstrates AI’s ability to produce “real assets” and accelerate pipelines.
• Strategic Partnerships: Large pharmaceutical companies are actively partnering with AI-native biotech startups, recognizing that AI is a critical capability they need to acquire or develop internally.
• India’s Position: Indian pharma companies are actively exploring and adopting AI. An EY report (Feb 2025) suggests that 50% of Indian Pharma companies are looking into AI-driven solutions, with a significant portion already using Generative AI. This indicates that the “when” is already upon the Indian pharmaceutical industry as it seeks to move beyond generics and into novel drug discovery.

4. To Leverage Data Explosion (Continuous Requirement):

• The sheer volume of biological, chemical, and clinical data being generated globally (genomics, proteomics, EHRs, real-world evidence) is overwhelming for human analysis. AI is required continuously to make sense of this “big data” and extract actionable insights.

In essence, AI-powered drug discovery is required now, and increasingly so, across every stage of the pharmaceutical value chain – from initial target identification and lead discovery to preclinical testing, clinical trial optimization, and even drug repurposing. It’s the critical technology enabling the industry to develop more effective, safer, and affordable medicines more rapidly, fundamentally transforming how new therapies reach patients.

Where is require AI-Powered Drug Discovery – Using AI to identify potential pharmaceutical compounds?

AI-powered drug discovery is being applied and is required in various locations and contexts across the globe, with a rapidly increasing footprint in India. It’s not confined to a single geographical “where” but rather to the types of institutions and organizations involved in pharmaceutical research and development.

Here’s where AI-powered drug discovery is required:

1. Major Pharmaceutical Hubs Globally:

• North America (Especially the United States): This is currently the leading region for AI in drug discovery. The presence of numerous large pharmaceutical companies, a strong biotech startup ecosystem, significant venture capital investment, and leading AI research institutions makes it a primary hub. Cities like Boston (Cambridge), San Francisco Bay Area, and San Diego are hotbeds of activity.
• Europe: Countries like the UK, Germany, Switzerland, and France have robust pharmaceutical industries and are actively fostering AI integration. Collaborations between pharma giants and AI tech providers are common.
• Asia-Pacific (APAC): This region is rapidly emerging as a significant player.
  • China: Investing heavily in AI for drug discovery, with numerous startups and government-backed initiatives.
  • India: A major focus given its strong pharmaceutical manufacturing base and increasing shift towards novel drug development.
  • Singapore, South Korea, Japan: Also making significant strides with strong academic institutions and government support.

2. Within Large Pharmaceutical Companies (Globally and in India):

• In-house R&D Departments: Major pharmaceutical companies are building their own AI capabilities, hiring data scientists and computational chemists, and establishing dedicated AI/ML divisions within their R&D centers.
• Partnership Offices: They are actively seeking out and partnering with AI-native biotech companies and AI software providers. These partnerships are often established globally but can have specific operational centers in various countries.
  • Indian Examples: Companies like Sun Pharma, Dr. Reddy’s Laboratories, Lupin, Cipla, Biocon, and Aurobindo Pharma are actively investing in or exploring AI-driven solutions to accelerate drug discovery and optimize operations. An EY report (Feb 2025) highlighted that 50% of Indian Pharma companies are exploring or investing in AI-driven solutions, with 25% already having Generative AI applications in production.

3. Biotechnology Startups and AI-Native Drug Discovery Companies:

• These companies are often born out of university research or by entrepreneurs explicitly focused on leveraging AI/ML as their core technology platform for drug discovery. They can be found in major tech and biotech clusters worldwide.
• Indian Context: India has a burgeoning ecosystem of biotech startups that are leveraging AI. While specific names might evolve rapidly, these companies are crucial players. Examples include Innoplexus (an Indo-German company with a significant presence in India focusing on AI for drug development), and other emerging startups in computational biology and cheminformatics.

4. Academic and Research Institutions:

• University Research Labs: Leading universities and research institutes globally are conducting foundational and applied research in AI for drug discovery. They are crucial for developing new algorithms, understanding complex biological systems, and training the next generation of scientists.
  • Indian Examples: Premier institutions like the Indian Institutes of Technology (IITs), the Indian Institute of Science (IISc), the National Institute of Pharmaceutical Education and Research (NIPERs), and various medical colleges and research institutes (e.g., THSTI, AIIMS) are involved in research at the intersection of AI, computational chemistry, and biology. NPTEL offers courses on AI in drug discovery, indicating academic engagement.
• Government Research Bodies: Organizations like the Indian Council of Medical Research (ICMR) are actively using AI to accelerate drug development, as seen in their collaboration for TB treatment research.

5. Contract Research Organizations (CROs) and Consultancies:

• CROs are increasingly adopting AI to enhance their services for pharmaceutical clients, particularly in areas like preclinical testing optimization, clinical trial design, and patient recruitment.
• Consulting firms specializing in life sciences and technology are advising and implementing AI solutions for their pharma and biotech clients.

In Summary:

AI-powered drug discovery is being applied and required wherever cutting-edge pharmaceutical research and development is taking place. This spans:

• Geographically diverse locations: North America, Europe, and increasingly Asia-Pacific, with India playing a pivotal role.
• Across different organizational structures: From established multinational pharmaceutical corporations and agile biotechnology startups to leading academic research institutions and specialized CROs.

The “where” is essentially any place that wants to be at the forefront of medical innovation and accelerate the delivery of new, life-saving therapies to patients.

Case study on AI-Powered Drug Discovery – Using AI to identify potential pharmaceutical compounds?

Courtesy: SandboxAQ

AI-powered drug discovery is generating a wealth of exciting case studies, demonstrating its ability to accelerate processes, reduce costs, and identify novel compounds that might otherwise be missed. Here are a few prominent examples, including a notable success from a company pioneering AI in drug discovery:

---

Case Study 1: Insilico Medicine – From AI-Designed to Clinical Trial in Record Time

Context: Insilico Medicine, a clinical-stage generative AI-driven drug discovery company, is a leading example of an AI-first approach. Their entire pipeline, from target discovery to molecule design, is powered by their proprietary AI platform, Pharma.AI. The traditional drug discovery process for a novel drug can take 10-15 years.

Problem Addressed: The immense time, cost, and high failure rates associated with traditional drug discovery, particularly for novel targets and compounds.

AI-Powered Solution and Process:

1. AI-Driven Target Discovery (PandaOmics): Insilico’s AI platform, specifically the “PandaOmics” module, analyzes vast biological and clinical datasets (genomic, proteomic, transcriptomic, clinical trial data, scientific literature) to identify novel disease targets that are strongly implicated in specific conditions. For example, for Idiopathic Pulmonary Fibrosis (IPF), their AI identified a novel target (related to fibrosis) that traditional methods might have overlooked.
2. Generative AI for Molecule Design (Chemistry42): Once a promising target is identified, Insilico’s “Chemistry42” generative AI platform steps in. This AI can:
   • De Novo Design: Generate entirely new molecular structures that are specifically designed to interact with the identified target. It considers desired properties like potency, selectivity, and drug-likeness (e.g., solubility, permeability, metabolic stability).
   • Rapid Optimization: Instead of synthesizing and testing thousands of compounds, Chemistry42 can iterate rapidly, predicting the best modifications to improve a lead compound’s properties, significantly reducing the “design-make-test-learn” cycle time. Insilico reports needing to synthesize and test only about 60-115 molecules on average per project, a fraction of what traditional methods require.
3. Predictive Preclinical Assessment: AI models are used to predict ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) properties and other crucial preclinical parameters, helping to filter out compounds with high predicted toxicity or poor pharmacokinetic profiles before expensive wet-lab testing.

Key Achievements and Impact:

• Accelerated Timeline to Clinic: Insilico Medicine became famous for putting an AI-discovered target and AI-designed drug, INS018_055 (for Idiopathic Pulmonary Fibrosis – IPF), into Phase II clinical trials in remarkably short timeframes.
  • The entire journey from target identification to entering Phase I clinical trials for INS018_055 took under 30 months, roughly half the industry average for a novel drug.
  • Subsequently, it progressed to Phase II trials, marking a significant milestone for AI-driven drug discovery by demonstrating a “first-in-class” AI-discovered and AI-designed drug in patient trials for a chronic disease.
• Multiple Clinical Candidates: As of late 2024 / early 2025, Insilico has nominated over 20 preclinical candidates and received IND (Investigational New Drug) clearance for 10 AI-enabled drug candidates to enter clinical trials across various therapeutic areas including fibrosis, cancer, and inflammatory bowel disease. This demonstrates a consistent, repeatable process enabled by AI.
• Efficiency: The company reports that its AI platform enables them to progress from project initiation to preclinical candidate nomination in an average of 12-18 months.

Learning Points:

• End-to-End AI Integration: Insilico’s success highlights the power of integrating AI across the entire drug discovery pipeline, rather than using it as a fragmented tool.
• Generative AI’s Potential: The ability of AI to design novel molecules, not just screen existing ones, is a game-changer.
• Validation through Clinical Progress: The most significant validation for AI in drug discovery comes from candidates successfully entering and progressing through clinical trials.

---

Case Study 2: BenevolentAI – Rapid Drug Repurposing for COVID-19

Context: BenevolentAI is a UK-based company that uses an AI-driven knowledge graph to accelerate drug discovery, particularly in areas like drug repurposing and target identification.

Problem Addressed: The urgent global need to find effective treatments for COVID-19 in early 2020, requiring rapid identification of existing drugs that could be repurposed.

AI-Powered Solution and Process:

1. AI-Enhanced Knowledge Graph: BenevolentAI’s core asset is a vast knowledge graph that computationally integrates and analyzes millions of scientific papers, genomic data, clinical trial results, and other biomedical information using AI (including NLP). This graph uncovers hidden relationships between diseases, genes, proteins, and existing drugs.
2. Rapid Hypothesis Generation: In early 2020, BenevolentAI’s scientists posed specific questions to their AI platform regarding COVID-19. The AI swiftly analyzed its knowledge graph to identify approved drugs that could potentially act against the novel coronavirus by affecting both the virus’s entry into cells and the body’s inflammatory response.
3. Identification of Baricitinib: Within days, the AI system identified baricitinib, an existing rheumatoid arthritis drug, as a potential candidate. The AI’s novel insight was that baricitinib could not only reduce inflammation (a key issue in severe COVID-19) but also inhibit a protein (AAK1) involved in the virus’s entry into lung cells.

Impact:

• Accelerated Clinical Trials: BenevolentAI published its findings in The Lancet in February 2020. This AI-driven hypothesis quickly led to clinical trials of baricitinib for COVID-19 by Eli Lilly.
• Emergency Use Authorization: Baricitinib, in combination with remdesivir, subsequently received Emergency Use Authorization (EUA) from the FDA in November 2020 and was approved for emergency use in various countries, including India.
• Proof of Concept for Repurposing: This case demonstrated the immense power of AI in rapidly identifying new indications for existing drugs, significantly faster and cheaper than developing new ones from scratch, crucial during a pandemic.

Learning Points:

• Power of Knowledge Graphs: AI-driven knowledge graphs are invaluable for extracting hidden insights from vast, unstructured biomedical data.
• Speed in a Crisis: AI can dramatically accelerate the process of finding therapeutic options, particularly in urgent public health scenarios.
• Drug Repurposing Efficiency: AI makes drug repurposing a highly efficient strategy for rapid drug development.

---

These case studies exemplify how AI is moving beyond academic curiosity to deliver tangible, impactful results in drug discovery, from de novo design to rapid repurposing, ultimately bringing new therapies to patients faster.

White paper on AI-Powered Drug Discovery – Using AI to identify potential pharmaceutical compounds?

White Paper: AI-Powered Drug Discovery – Revolutionizing Pharmaceutical Innovation in India

---

Executive Summary

The global pharmaceutical landscape is at an inflection point, driven by an urgent need for accelerated drug development and reduced R&D costs. India, with its well-established pharmaceutical industry and burgeoning tech talent, stands poised to lead this transformation. The traditional drug discovery paradigm, characterized by its extensive timelines (10-15 years), astronomical costs (billions of dollars per drug), and high failure rates (over 90% in clinical trials), is no longer sustainable for addressing the complex health challenges of the 21st century.

Artificial Intelligence (AI) is emerging as the most disruptive technology in drug discovery, offering unprecedented capabilities to analyze vast datasets, predict molecular interactions, design novel compounds, and optimize clinical trials with remarkable speed and precision. This white paper delves into the transformative potential of AI in identifying potential pharmaceutical compounds, outlines its core mechanisms, highlights its industrial applications and growing adoption in India, and addresses the critical regulatory and ethical considerations. We present a strategic roadmap for India to leverage AI as a cornerstone of its pharmaceutical innovation ecosystem, solidifying its position as a global leader in accessible and affordable healthcare solutions.

1. The Inevitable Shift: Why AI is Indispensable for Drug Discovery

The pharmaceutical industry faces immense pressure to innovate faster and more efficiently. The “Eroom’s Law” (Moore’s Law backward) illustrates that the cost of bringing a new drug to market doubles approximately every nine years. This unsustainable trend, coupled with the rising complexity of diseases and the emergence of new pathogens, demands a radical shift in R&D strategy.

AI offers a compelling solution by:

• Handling Big Data: The explosion of biomedical data (genomics, proteomics, chemical libraries, clinical trial results, real-world evidence) is beyond human comprehension. AI algorithms can process, integrate, and derive insights from this massive, multi-modal data.
• Accelerating Timelines: AI can compress years of traditional research into months or even weeks, particularly in early discovery phases.
• Reducing Costs & Failure Rates: By predicting efficacy, toxicity, and other critical properties early in the pipeline, AI can identify problematic candidates sooner, preventing costly late-stage failures.
• Unlocking Novelty: Generative AI can design entirely new molecules with desired properties, expanding the search beyond known chemical spaces.
• Enabling Precision Medicine: AI analyzes patient-specific data to identify optimal drug candidates and personalize treatment strategies.

2. Core Mechanisms: How AI Identifies Potential Pharmaceutical Compounds

AI’s power in drug discovery stems from its ability to learn from data and make predictions. Key AI methodologies employed include:

2.1. Machine Learning (ML): ML algorithms are trained on existing data to identify patterns and make predictions.

• Supervised Learning: Used for tasks where labeled data is available (e.g., predicting if a compound is toxic based on known toxic/non-toxic compounds).
• Unsupervised Learning: Used for identifying hidden patterns in unlabeled data (e.g., clustering compounds with similar properties).
• Reinforcement Learning: AI agents learn to make sequences of decisions (e.g., in optimizing molecular structures) through trial and error, based on rewards and penalties.

2.2. Deep Learning (DL): A subset of ML using neural networks with multiple layers, highly effective for complex pattern recognition.

• Convolutional Neural Networks (CNNs): Excellent for processing grid-like data, such as images (e.g., analyzing cell images for drug effects) and molecular representations.
• Recurrent Neural Networks (RNNs) & Transformers: Ideal for sequential data like protein sequences or chemical strings (SMILES notation), used in generative chemistry.
• Graph Neural Networks (GNNs): Particularly powerful for modeling molecular structures as graphs (atoms as nodes, bonds as edges), capturing complex relationships.

2.3. Natural Language Processing (NLP): Enables AI to understand, interpret, and generate human language.

• Automated Literature Mining: Extracting insights from millions of scientific papers, patents, and clinical trial reports to build knowledge bases and identify new targets or drug repurposing opportunities.
• Building Knowledge Graphs: Creating interconnected networks of biological entities (genes, proteins, diseases), compounds, and their relationships, allowing AI to query and discover novel links.

2.4. Generative AI: A cutting-edge branch of AI that can create new data (e.g., novel molecules) rather than just analyze existing data.

• Generative Adversarial Networks (GANs) & Variational Autoencoders (VAEs): Used to design entirely new molecular structures with specific desired properties (e.g., binding affinity, low toxicity) tailored to a particular disease target.

3. Industrial Applications: AI Across the Drug Discovery Pipeline

AI’s impact is transformative at every stage of pharmaceutical R&D:

3.1. Target Identification and Validation:

• Mechanism: AI analyzes vast multi-omics datasets (genomics, proteomics, transcriptomics) to identify disease-relevant genes and proteins, and predicts their 3D structures (e.g., AlphaFold’s impact). It also mines scientific literature for overlooked targets.
• Impact: Pinpoints the most promising biological targets, reducing the time and cost of early-stage research and focusing efforts on high-potential avenues.

3.2. Lead Discovery and Hit Identification (Virtual Screening):

• Mechanism: AI computationally screens billions of chemical compounds against a target’s structure (molecular docking simulations) to predict binding affinity. Generative AI designs novel molecules from scratch, optimizing for desired properties.
• Impact: Drastically reduces the need for costly and time-consuming physical high-throughput screening, accelerating the identification of initial “hits” and novel compounds.

3.3. Lead Optimization and Property Prediction:

• Mechanism: AI models predict critical ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) properties and other physiochemical characteristics of lead compounds. It guides medicinal chemists in making precise structural modifications to enhance potency, selectivity, and safety.
• Impact: Minimizes the number of compounds that need to be synthesized and tested, streamlining the optimization process and increasing the likelihood of successful preclinical development.

3.4. Drug Repurposing (Repositioning):

• Mechanism: AI analyzes existing drug databases, clinical trial results, and real-world evidence to identify new therapeutic uses for already approved drugs.
• Impact: A significantly faster, less risky, and more cost-effective path to new therapies, as repurposed drugs have already undergone extensive safety testing. Proven invaluable during the COVID-19 pandemic.

3.5. Preclinical and Clinical Trial Optimization:

• Mechanism: AI assists with patient stratification and recruitment (identifying ideal candidates based on genetic profiles), optimizes trial design, predicts potential adverse events, and analyzes trial data in real-time.
• Impact: Reduces the high failure rates of clinical trials, shortens trial timelines, lowers overall development costs, and facilitates the delivery of personalized medicines.

4. The Indian Imperative: AI in India’s Pharmaceutical Landscape

India, often dubbed the “Pharmacy of the World” due to its robust generics manufacturing capabilities, is now strategically shifting towards novel drug discovery. AI is the key enabler for this transition.

• Growing Adoption: An EY-Parthenon & Microsoft report from February 2025 highlights that 50% of Indian pharmaceutical companies are already exploring or investing in AI-driven solutions, with 25% already having Generative AI applications in production. This signifies a strong and accelerating adoption curve.
• Cost and Time Savings: For a nation focused on affordable healthcare, AI’s ability to reduce drug development costs and timelines is a significant draw, enabling the creation of novel therapies at a more accessible price point.
• Abundant Talent Pool: India boasts a large and skilled workforce in IT, data science, and biotechnology, providing a fertile ground for developing and deploying AI solutions in drug discovery.
• Government Support: The Indian government’s emphasis on “Digital India” and its proactive approach to AI governance (e.g., National Strategy for AI, AI for All initiative, and Union Budget 2025-26’s focus on AI in healthcare and drug discovery) is fostering a supportive ecosystem. The Indian Council of Medical Research (ICMR) is actively leveraging AI for accelerating drug discovery, particularly for diseases like Tuberculosis.
• Data Availability: India’s vast and diverse patient population generates rich clinical data, which, if properly anonymized and leveraged, can be a powerful asset for training robust AI models for personalized medicine.
• Emergence of Indian AI-Biotech Startups: A growing number of Indian startups are emerging with AI-first approaches to drug discovery, often collaborating with established pharma players or academic institutions.

5. Navigating Challenges: Regulatory and Ethical Considerations in India

While AI offers immense promise, its responsible deployment is critical. India’s evolving regulatory landscape is key:

• Data Privacy (Digital Personal Data Protection Act, 2023): AI in drug discovery relies on vast datasets, including sensitive patient information. Adherence to the DPDP Act’s principles of consent, data minimization, purpose limitation, and robust security safeguards is paramount.
• Ethical AI (MeitY’s AI Governance Guidelines): Principles of Fairness, Transparency, Accountability, Privacy, Security, Safety, Reliability, and Robustness must guide AI model development.
  • Bias Mitigation: Ensuring AI models are not biased against certain demographic groups, which could lead to health inequities.
  • Explainable AI (XAI): As AI becomes more complex (“black box” models), understanding why an AI made a certain prediction is crucial for scientific validation, regulatory approval, and accountability.
  • Adversarial AI: Protecting AI models themselves from malicious attacks (e.g., data poisoning) that could compromise their integrity and lead to erroneous drug discovery.
• Regulatory Adaptation: Indian regulatory bodies (CDSCO, ICMR) must continue to evolve their guidelines to accommodate AI-driven drug development, ensuring safety and efficacy while facilitating innovation.
• Interoperability and Data Standards: Harmonizing data formats and ensuring interoperability across different data sources (e.g., Electronic Health Records) is crucial for building comprehensive AI training datasets.

6. Strategic Recommendations for India

To fully harness the potential of AI-powered drug discovery and establish India as a global leader, the following actions are recommended:

• Invest in National AI Computing Infrastructure: Develop and provide access to high-performance computing resources and cloud-based AI platforms essential for complex drug discovery simulations.
• Foster Collaborative Data Ecosystems: Create secure, privacy-preserving frameworks for data sharing between academia, industry, and healthcare providers to build large, diverse, and high-quality datasets for AI training.
• Develop Adaptive Regulatory Frameworks: Establish clear, agile, and forward-looking regulations for AI in drug discovery, focusing on validation, explainability, and responsible deployment. Participate actively in global discussions on AI regulation in healthcare.
• Prioritize Talent Development: Launch aggressive national programs to skill and re-skill professionals in AI, machine learning, bioinformatics, and computational chemistry. Encourage interdisciplinary research and education.
• Incentivize R&D and Startups: Provide targeted grants, funding, and tax incentives for AI-driven drug discovery research and for biotech startups leveraging AI as their core technology.
• Promote Public-Private Partnerships (PPPs): Facilitate collaborations between government research bodies (e.g., ICMR, NIPERs), large pharmaceutical companies, and AI-native startups to accelerate innovation and technology transfer.
• Focus on Specific Disease Areas: Leverage AI to address high-burden diseases in India (e.g., tuberculosis, diabetes, neglected tropical diseases) to demonstrate tangible impact and develop affordable solutions.
• Ethical AI Governance: Develop robust ethical guidelines and best practices for AI deployment in healthcare, ensuring fairness, transparency, and patient safety.

7. Conclusion

AI-powered drug discovery is not merely an incremental improvement; it is a paradigm shift that promises to redefine how new medicines are conceived, developed, and delivered to patients. For India, this revolution represents an unparalleled opportunity to transcend its role as a generics powerhouse and emerge as a global leader in novel, affordable pharmaceutical innovation. By strategically investing in AI infrastructure, fostering collaboration, nurturing talent, and developing a supportive regulatory environment, India can unlock the full potential of AI to accelerate drug discovery, address critical health challenges, and secure its position at the forefront of global healthcare. The future of medicine is intelligent, and India is poised to lead the way.

Industrial Application of AI-Powered Drug Discovery – Using AI to identify potential pharmaceutical compounds?

AI-powered drug discovery is no longer a theoretical concept but a practical, industrial application that is rapidly transforming the pharmaceutical and biotechnology sectors. Companies are leveraging AI at various stages of drug development to accelerate timelines, reduce costs, and improve success rates.

Here are the key industrial applications of AI-powered drug discovery:

1. Automated Target Identification and Validation

• Traditional Approach: Identifying suitable biological targets (e.g., proteins, genes, pathways) that are implicated in a disease and can be modulated by a drug is a complex and often slow process, requiring extensive experimental validation.
• AI Application:
  • Big Data Analysis: AI systems (using machine learning and natural language processing) analyze vast amounts of “omics” data (genomics, proteomics, transcriptomics, metabolomics), patient data, electronic health records (EHRs), and millions of scientific publications and patents.
  • Pattern Recognition: They identify subtle patterns, correlations, and causal relationships that indicate which targets are most likely to be effective for a specific disease.
  • Knowledge Graphs: AI constructs and queries “knowledge graphs” that link genes, proteins, diseases, compounds, and pathways, allowing researchers to uncover novel or under-explored therapeutic targets.
  • Protein Structure Prediction: AI models like AlphaFold predict the highly accurate 3D structures of proteins, which is crucial for understanding how a drug can bind to and interact with a target.
• Industrial Impact: Significantly reduces the time and resources spent on the initial, critical step of drug discovery, allowing companies to focus on the most promising targets.

2. High-Throughput Virtual Screening and Lead Discovery

• Traditional Approach: Physical high-throughput screening (HTS) involves testing millions of chemical compounds in a laboratory against a biological target. This is expensive, time-consuming, and limited by the size of physical compound libraries.
• AI Application:
  • Virtual Screening: AI algorithms (e.g., deep learning models, molecular docking simulations) computationally screen billions of virtual compounds from enormous databases against the 3D structure of a target protein. They predict binding affinity and potential interactions.
  • Filtering and Prioritization: AI rapidly filters out compounds unlikely to be effective or those with predicted undesirable properties (like toxicity), guiding experimental chemists to synthesize and test only the most promising “hits.”
  • De Novo Design (Generative AI): Generative AI models (e.g., GANs, VAEs) can design completely new chemical compounds from scratch that have never existed before. Researchers specify desired properties (e.g., high potency, specific ADMET profile), and the AI generates novel molecular structures tailored to fit the target’s binding site.
• Industrial Impact: Dramatically shortens the hit identification phase from years to months, reduces laboratory costs, and unlocks the ability to explore a much larger chemical space for novel drug candidates.

3. Lead Optimization and ADMET Prediction

• Traditional Approach: Once initial “hits” are found, they need to be optimized for various “drug-like” properties such as potency, selectivity, solubility, permeability, metabolic stability, and low toxicity (ADMET properties). This is an iterative and often unpredictable process for medicinal chemists.
• AI Application:
  • Predictive Modeling: AI models are trained on vast datasets of known compound properties. They can accurately predict the ADMET profile of a new or modified compound based on its chemical structure, guiding optimization efforts.
  • Structure-Activity Relationship (SAR) Prediction: AI identifies complex relationships between a compound’s chemical structure and its biological activity, helping chemists understand how to modify a molecule to improve its desired characteristics.
  • Automated Design of Experiments: AI can suggest optimal sets of experiments to run to gather the most informative data for lead optimization.
• Industrial Impact: Accelerates the lead optimization cycle, reduces the number of compounds that need to be synthesized and tested, and improves the chances of a candidate succeeding in later development stages by identifying potential issues early.

4. Drug Repurposing (Repositioning)

• Traditional Approach: Finding new therapeutic uses for existing, approved drugs (or those that failed for their original indication) is often serendipitous or requires extensive manual literature review.
• AI Application:
  • Knowledge Graph Mining: AI analyzes its comprehensive knowledge graphs, patient data, and clinical trial results to uncover hidden connections between existing drugs, their mechanisms of action, and new disease indications.
  • Systematic Identification: The AI can systematically screen all approved drugs against a specific disease pathway or a new pathogen (as seen with COVID-19) to identify promising candidates for repurposing.
• Industrial Impact: Offers a faster, less risky, and more cost-effective path to new therapies since repurposed drugs have already passed significant safety hurdles.

5. Clinical Trial Design and Optimization

• Traditional Approach: Clinical trials are the most expensive, time-consuming, and highest-risk phase of drug development, with challenges in patient recruitment, site selection, and predicting outcomes.
• AI Application:
  • Patient Stratification and Recruitment: AI analyzes patient genomics, medical history, real-world evidence, and demographic data to identify ideal patient populations for trials, accelerating recruitment and improving trial success rates.
  • Predictive Analytics for Outcomes: AI can predict the likelihood of success, potential adverse events, or optimal dosing regimens based on preclinical data and historical trial information.
  • Biomarker Discovery: AI identifies biomarkers that can predict a patient’s response to a drug or disease progression, leading to more targeted therapies and personalized medicine.
• Industrial Impact: Reduces clinical trial failure rates, shortens trial timelines, lowers overall development costs, and facilitates the development of more effective, personalized treatments.

Examples in the Indian Context:

While specific proprietary details are often confidential, Indian pharmaceutical companies and emerging biotech firms are actively investing in these areas:

• Major Pharma Players: Companies like Sun Pharma, Dr. Reddy’s Laboratories, Cipla, Lupin, Biocon, and Aurobindo Pharma are integrating AI into their R&D, focusing on areas like predictive analytics for manufacturing efficiency, supply chain optimization, and increasingly, drug discovery. An EY report (Feb 2025) indicated 50% of Indian Pharma companies are exploring or investing in AI-driven solutions, with 25% already having Generative AI applications in production.
• Emerging AI-Biotech Startups: Several Indian startups are specializing in AI for drug discovery, leveraging the country’s strong tech and scientific talent pool. These companies often focus on specific therapeutic areas or AI methodologies.
• Academic Collaborations: Indian research institutions (IITs, IISc, NIPERs) are engaging in AI-driven drug discovery research, often collaborating with industry partners to translate academic findings into industrial applications.

In conclusion, AI-powered drug discovery is fundamentally reshaping the pharmaceutical industry by providing unprecedented capabilities for speed, accuracy, and innovation. Its industrial applications are leading to more efficient R&D, reduced costs, and ultimately, faster delivery of new and more effective medicines to patients worldwide, including in India where its adoption is rapidly gaining momentum.

1.  Decision trees: Russell & Norvig (2021, sect. 19.3), Domingos (2015, p. 88)
2. ^Non-parameteric learning models such as K-nearest neighbor and support vector machines: Russell & Norvig (2021, sect. 19.7), Domingos (2015, p. 187) (k-nearest neighbor)
   • Domingos (2015, p. 88) (kernel methods)
3. ^ Domingos (2015), p. 152.
4. ^ Naive Bayes classifier: Russell & Norvig (2021, sect. 12.6), Domingos (2015, p. 152)
5. ^ Jump up to:^a b Neural networks: Russell & Norvig (2021, chpt. 21), Domingos (2015, Chapter 4)
6. ^ Gradient calculation in computational graphs, backpropagation, automatic differentiation: Russell & Norvig (2021, sect. 21.2), Luger & Stubblefield (2004, pp. 467–474), Nilsson (1998, chpt. 3.3)
7. ^ Universal approximation theorem: Russell & Norvig (2021, p. 752) The theorem: Cybenko (1988), Hornik, Stinchcombe & White (1989)
8. ^ Feedforward neural networks: Russell & Norvig (2021, sect. 21.1)
9. ^ Recurrent neural networks: Russell & Norvig (2021, sect. 21.6)
10. ^ Perceptrons: Russell & Norvig (2021, pp. 21, 22, 683, 22)
11. ^ Jump up to:^a b Deep learning: Russell & Norvig (2021, chpt. 21), Goodfellow, Bengio & Courville (2016), Hinton et al. (2016), Schmidhuber (2015)
12. ^ Convolutional neural networks: Russell & Norvig (2021, sect. 21.3)
13. ^ Sindhu V, Nivedha S, Prakash M (February 2020). “An Empirical Science Research on Bioinformatics in Machine Learning”. Journal of Mechanics of Continua and Mathematical Sciences (7). doi:10.26782/jmcms.spl.7/2020.02.00006.
14. ^ Deng & Yu (2014), pp. 199–200.
15. ^ Ciresan, Meier & Schmidhuber (2012).
16. ^ Russell & Norvig (2021), p. 750.
17. ^ Jump up to:^a b c Russell & Norvig (2021), p. 17.
18. ^ Jump up to:^{a b c d e f g} Russell & Norvig (2021), p. 785.
19. ^ Jump up to:^a b Schmidhuber (2022), sect. 5.
20. ^ Schmidhuber (2022), sect. 6.
21. ^ Jump up to:^a b c Schmidhuber (2022), sect. 7.
22. ^ Schmidhuber (2022), sect. 8.
23. ^ Quoted in Christian (2020, p. 22)
24. ^ Metz, Cade; Weise, Karen (5 May 2025). “A.I. Hallucinations Are Getting Worse, Even as New Systems Become More Powerful”. The New York Times. ISSN 0362-4331. Retrieved 6 May 2025.
25. ^ Smith (2023).
26. ^ “Explained: Generative AI”. 9 November 2023.
27. ^ “AI Writing and Content Creation Tools”. MIT Sloan Teaching & Learning Technologies. Archived from the original on 25 December 2023. Retrieved 25 December 2023.
28. ^ Marmouyet (2023).
29. ^ Kobielus (2019).
30. ^ Thomason, James (21 May 2024). “Mojo Rising: The resurgence of AI-first programming languages”. VentureBeat. Archived from the original on 27 June 2024. Retrieved 26 May 2024.
31. ^ Wodecki, Ben (5 May 2023). “7 AI Programming Languages You Need to Know”. AI Business. Archived from the original on 25 July 2024. Retrieved 5 October 2024.
32. ^ Plumb, Taryn (18 September 2024). “Why Jensen Huang and Marc Benioff see ‘gigantic’ opportunity for agentic AI”. VentureBeat. Archived from the original on 5 October 2024. Retrieved 4 October 2024.
33. ^ Mims, Christopher (19 September 2020). “Huang’s Law Is the New Moore’s Law, and Explains Why Nvidia Wants Arm”. Wall Street Journal. ISSN 0099-9660. Archived from the original on 2 October 2023. Retrieved 19 January 2025.
34. ^ Davenport, T; Kalakota, R (June 2019). “The potential for artificial intelligence in healthcare”. Future Healthc J. 6 (2): 94–98. doi:10.7861/futurehosp.6-2-94. PMC 6616181. PMID 31363513.
35. ^ Lyakhova, U.A.; Lyakhov, P.A. (2024). “Systematic review of approaches to detection and classification of skin cancer using artificial intelligence: Development and prospects”. Computers in Biology and Medicine. 178: 108742. doi:10.1016/j.compbiomed.2024.108742. PMID 38875908. Archived from the original on 3 December 2024. Retrieved 10 October 2024.
36. ^ Alqudaihi, Kawther S.; Aslam, Nida; Khan, Irfan Ullah; Almuhaideb, Abdullah M.; Alsunaidi, Shikah J.; Ibrahim, Nehad M. Abdel Rahman; Alhaidari, Fahd A.; Shaikh, Fatema S.; Alsenbel, Yasmine M.; Alalharith, Dima M.; Alharthi, Hajar M.; Alghamdi, Wejdan M.; Alshahrani, Mohammed S. (2021). “Cough Sound Detection and Diagnosis Using Artificial Intelligence Techniques: Challenges and Opportunities”. IEEE Access. 9: 102327–102344. Bibcode:2021IEEEA…9j2327A. doi:10.1109/ACCESS.2021.3097559. ISSN 2169-3536. PMC 8545201. PMID 34786317.
37. ^ Jump up to:^a b Bax, Monique; Thorpe, Jordan; Romanov, Valentin (December 2023). “The future of personalized cardiovascular medicine demands 3D and 4D printing, stem cells, and artificial intelligence”. Frontiers in Sensors. 4. doi:10.3389/fsens.2023.1294721. ISSN 2673-5067.
38. ^ Dankwa-Mullan, Irene (2024). “Health Equity and Ethical Considerations in Using Artificial Intelligence in Public Health and Medicine”. Preventing Chronic Disease. 21: E64. doi:10.5888/pcd21.240245. ISSN 1545-1151. PMC 11364282. PMID 39173183.
39. ^ Jumper, J; Evans, R; Pritzel, A (2021). “Highly accurate protein structure prediction with AlphaFold”. Nature. 596 (7873): 583–589. Bibcode:2021Natur.596..583J. doi:10.1038/s41586-021-03819-2. PMC 8371605. PMID 34265844.
40. ^ “AI discovers new class of antibiotics to kill drug-resistant bacteria”. 20 December 2023. Archived from the original on 16 September 2024. Retrieved 5 October 2024.
41. ^ “AI speeds up drug design for Parkinson’s ten-fold”. Cambridge University. 17 April 2024. Archived from the original on 5 October 2024. Retrieved 5 October 2024.
42. ^ Horne, Robert I.; Andrzejewska, Ewa A.; Alam, Parvez; Brotzakis, Z. Faidon; Srivastava, Ankit; Aubert, Alice; Nowinska, Magdalena; Gregory, Rebecca C.; Staats, Roxine; Possenti, Andrea; Chia, Sean; Sormanni, Pietro; Ghetti, Bernardino; Caughey, Byron; Knowles, Tuomas P. J.; Vendruscolo, Michele (17 April 2024). “Discovery of potent inhibitors of α-synuclein aggregation using structure-based iterative learning”. Nature Chemical Biology. 20 (5). Nature: 634–645. doi:10.1038/s41589-024-01580-x. PMC 11062903. PMID 38632492.
43. ^ Grant, Eugene F.; Lardner, Rex (25 July 1952). “The Talk of the Town – It”. The New Yorker. ISSN 0028-792X. Archived from the original on 16 February 2020. Retrieved 28 January 2024.
44. ^ Anderson, Mark Robert (11 May 2017). “Twenty years on from Deep Blue vs Kasparov: how a chess match started the big data revolution”. The Conversation. Archived from the original on 17 September 2024. Retrieved 28 January 2024.
45. ^ Markoff, John (16 February 2011). “Computer Wins on ‘Jeopardy!’: Trivial, It’s Not”. The New York Times. ISSN 0362-4331. Archived from the original on 22 October 2014. Retrieved 28 January 2024.
46. ^ Byford, Sam (27 May 2017). “AlphaGo retires from competitive Go after defeating world number one 3–0”. The Verge. Archived from the original on 7 June 2017. Retrieved 28 January 2024.
47. ^ Brown, Noam; Sandholm, Tuomas (30 August 2019). “Superhuman AI for multiplayer poker”. Science. 365 (6456): 885–890. Bibcode:2019Sci…365..885B. doi:10.1126/science.aay2400. ISSN 0036-8075. PMID 31296650.
48. ^ “MuZero: Mastering Go, chess, shogi and Atari without rules”. Google DeepMind. 23 December 2020. Retrieved 28 January 2024.
49. ^ Sample, Ian (30 October 2019). “AI becomes grandmaster in ‘fiendishly complex’ StarCraft II”. The Guardian. ISSN 0261-3077. Archived from the original on 29 December 2020. Retrieved 28 January 2024.
50. ^ Wurman, P. R.; Barrett, S.; Kawamoto, K. (2022). “Outracing champion Gran Turismo drivers with deep reinforcement learning” (PDF). Nature. 602 (7896): 223–228. Bibcode:2022Natur.602..223W. doi:10.1038/s41586-021-04357-7. PMID 35140384.