Towards AI-designed Pipeline of Antifibrotic Therapies:
New Milestone in Kidney Fibrosis

Building on top of recent success in identifying a novel pan-fibrotic target and a promising preclinical candidate in Idiopathic Pulmonary Fibrosis (IPF) using AI-platform, Insilico Medicine reports a new milestone — an in vivo active molecule to tackle Kidney Fibrosis (KF) — a growing unmet medical need.

What is Fibrosis?
Our body has powerful repair mechanisms allowing for wound healing — a complex dynamic process involving a tightly regulated cascade of four physiological events, including coagulation, granulation tissue formation, re-epithelialization, and extracellular matrix (ECM) remodeling. Under certain unfavorable conditions, normal wound healing may transform into fibrosis — a pathological process, also known as fibrotic scarring. It is when normal functioning tissues in our body are gradually replaced by non-functional scar tissue, leading to a loss of function by organs and systems, and potentially — death.

According to one source published in 2008, 45% of all deaths in Western world were associated with some form of fibrosis. Risks of developing various fibroproliferative diseases, including idiopathic pulmonary fibrosis, liver cirrhosis, cardiovascular disease, and progressive kidney disease, are highly correlated with older age, which means fibrosis is not only a global health problem, but also a growing one — due to the increasingly ageing population.

When our epithelial or endothelial cells are damaged by mechanical impact, or as a result of disease, exposure to aggressive chemicals or microorganisms, the coagulation cascade releases special pro-inflammatory molecules — cytokines, which activate some immune cells, mainly neutrophils and macrophages. The job of activated immune cells is to remove tissue debris and dead cells, which leads to acute inflammation.

Meanwhile, immune cells themselves release a number of bioactive molecules (so-called "factors") — like chemokines and cytokines — to amplify inflammatory reactions. Next, the released factors, such as TGF-β, platelet derived growth factor (PDGF), interleukin-13 and interleukin-4, promote a controlled activation and proliferation of myofibroblasts — a special-purpose cells, which have structural functions and also muscle-like properties. Because of their muscle-like nature, myofibroblasts can generate cell traction force. During wound healing, myofibroblasts migrate to injury sites and render mechanical wound closure.

In the normal process, myofibroblasts render a balance between synthesis and degradation of the extracellular matrix (ECM) — a three-dimensional network consisting of extracellular macromolecules and minerals, which provides structural and biochemical support to surrounding cells — resulting in ECM homeostasis. At the final stage of normal wound healing, immune cells undergo apoptosis, while epithelial or endothelial cells proliferate to regenerate injury sites, leading to wound repair.

However, the paramount complexity of the wound healing process is also the cause of its fragility. Various factors, like diabetes, venous or arterial disease, infection, and age-related metabolic disruptions may lead to disruptions in normal repair routes. When stimulated persistently, pro-inflammatory cytokines or growth factors may become overexpressed and would overactivate various receptors. On the other hand, other biomolecules may suffer deficiencies. Both overexpression and deficiency can then lead to switching from normal wound healing to a malignant pro-fibrotic process. As a result, there might be the excess of immune cells, overly activated myofibroblast formation and proliferation, and abnormally high production rate of ECM.

Bad comes to worse, such pro-fibrotic process also additionally activates the pro-inflammatory factors, leading to further increase in inflammation and causing chronic inflammation.

Finally, continuous myofibroblasts activation would generate masses of ECM and shift the ECM homeostasis towards fibrosis. Pro-fibrotic process itself can be a cause of secondary injury to the wound and lead to a chronic vicious circle of pathological responses.

Being a systems disease, fibrosis may occur in various tissues and organs, but more often it develops in heart, lung, kidney, liver and skin. Fibrosis in different fibroproliferative diseases has both unique mechanisms, and also common similarities that regulate the process.

Researchers at Insilico Medicine have been focusing on common fibrosis mechanisms, and recently identified a novel pan-fibrotic target which promises to be a useful starting point for a broad range of antifibrotic drug discovery programs. A systems approach to antifibrotic drug discovery, enabled by a sophisticated deep learning (DL)-driven target identification platform PandaOmics, in concert with lead discovery platform Chemistry42, allowed Insilico Medicine to rapidly achieve drug design success with two fibrotic indications — development of novel preclinical drug candidate for IPF within under 18 month, and more recently a novel milestone in kidney fibrosis.
Kidney Fibrosis — a Growing Unmet Medical Need
The kidney has crucial roles in human physiology, including removal of waste products, maintaining balance of the body's fluids, involvement in metabolism and production of vitamin D. The kidney also produces hormones that regulate red blood cell production and blood pressure. A complex interplay of kidney's function with other organs and processes is the reason that kidney malfunctions have knock-on effects throughout the entire body and often lead to detrimental consequences and early death.

Kidney disease, also known as chronic kidney diseases or CKDs, causes more deaths than breast cancer or prostate cancer (NVS 2021 report of 2018 data). CKDs are a leading cause of death in the U.S. According to the National Kidney Foundation, CKS is an under-recognized public health crisis.

Renal fibrosis, or kidney fibrosis, is a major outcome of chronic kidney diseases, affecting up to 10% of the population and heavily impacting health systems worldwide. While CKD can be triggered by different factors, like diabetes, hypertension, chemical toxicity, or infection, its common result -- kidney fibrosis — gradually leads to complete renal dysfunction. At the early stage there might be no symptoms, so patients often get diagnosed with kidney fibrosis at late stages, when the only remaining options would be dialysis or kidney transplantation.

In 2016 there were more than 100.000 patients in the queue for kidney transplantation in the US alone, with over 3000 new patients being added to the list every month. On average, one in the queue has to spend around 3.6 years on dialysis until they can get a kidney transplant. The mortality of patients on dialysis can be as high as 20% per year and transplantation is limited by organ shortage. Even after transplantation, new organs gradually fail, typically within 10 years, requiring new dialysis or new transplantation.

The incidence of CKDs is highest among people older than 65, so the kidney disease crisis is a growing issue in the increasingly aging population and has to be addressed as soon as possible.
Challenges of Tackling Kidney Fibrosis
While billions of dollars have been poured into kidney disease research in the past decades, there has been no tangible progress in clinical practice related to efficient treatment of CKD and other pathologies. The number of randomized controlled trials (RCTs) published in nephrology is lower compared to other therapeutic areas, and most large RCTs in nephrology yield negative results. There is also a lack of nephrology studies evaluating hard clinical endpoints or surrogate endpoints.

The kidney is a highly complex organ, composed of a filter unit and a tubular segment, together containing over 20 different cell types. Since it is kidney fibrosis that is commonly believed to play a major role in the pathology of CKD and its progression to end-stage renal disease requiring dialysis and transplantation, many efforts have been focused on targeting fibrotic pathways and targets. Scientists focused heavily on major growth factors or cytokines that were early shown to drive both inflammation and fibrosis during progression of renal disease in animal models. Some popular examples include angiotensin II (Ang-II), transforming growth factor-β (TGF-β), platelet-derived growth factors (PDGFs), connective tissue growth factor (CTGF), endothelin-1 (ET-1), macrophage chemoattractant protein-1 (MCP-1), and tumor necrosis factor-α (TNF-α).

Despite the fact that some drug candidates targeting some of these factors made their ways into clinical trials, and inhibitors of the renin-angiotensin system are currently used in subgroups of CKD patients, the overall situation with targeting specifically kidney fibrosis is quite pessimistic.

Many drug discovery failures in the area of kidney fibrosis are due to poorly selected targets, or imperfect lead molecules. Also, renal fibrosis is mediated by a complex signaling network of several different pathways, therefore, targeting one pathway is not enough to be effective.

In order to discover efficient drugs for the kidney fibrosis, it is essential to have robust disease models and select pan-fibrotic targets that have substantial effect on the disease.
Deep Learning To The Rescue
Complexity of kidney fibrosis is paramount, and understanding disease mechanisms requires cutting-edge technologies and systems approaches. On the one hand, multi-omics analysis provides essential data about underlying molecular processes involved in fibrosis development and progression. The seamless combination of traditional transcriptomics approaches with emerging technologies, including proteomics, metabolomics, and metagenomics opened room for unprecedented opportunities to look inside biological systems.

On the other hand, the application of Artificial Intelligence, more specifically -- Deep Learning -- has proved to be beneficial at every step of the drug discovery process, especially at the hypothesis generation and target identification stage. Deep learning models and natural language processing technologies are compelling when modeling large complex multi-dimensional data sets, such as genomics, proteomics, clinical data, structural data about targets, and unstructured text (research papers, patents, grants etc.).
The ability to process multimodal big data and build sophisticated disease models for target selection was the key advantage of Insilico Medicine, when the company decided to enter the fibrosis drug discovery space with its AI-platforms PandaOmics and Chemistry42. The platforms include hundreds of modules such as generative adversarial neural nets (GANs), natural language processing (NLP) engines, and statistical components — all working in concert.
Our team spent years building and integrating hundreds of AI models, each responsible for a particular task, into one platform that is able to generate hypotheses, select targets, generate compounds, and predict clinical trial outcomes. To our knowledge, this is the most comprehensive drug design platform on the market.

In 2020 our team achieved the first major proof-of-concept success in the area of fibrosis using our integrated drug discovery platform, based on generative adversarial networks (GANs). As a result, we discovered a novel intracellular pan-fibrotic target, and designed a drug candidate that showed outstanding in vivo results for Idiopathic Pulmonary Fibrosis.

To build our initial hypothesis, we trained deep neural networks on a collection of omics and clinical datasets to predict tissue-specific fibrosis, as well as age and sex of the patients. We then applied a range of target discovery tools implemented in our PandaOmics target discovery system and performed sophisticated gene and pathway scoring published in Nature Communications, and came up with relevant targets via deep feature selection, causality inference, and de novo pathway reconstruction. The target novelty and disease association scoring was assessed by a natural language processing (NLP) engine, which analyses data from millions of data files, including patents, research publications, grants, and databases of clinical trials. As a result, PandaOmics revealed 20 targets for validation that we narrowed down to one novel intracellular target and prioritized for further analysis.

Chemistry42 is our generative chemistry module for drug discovery. This module includes an ensemble of generative and scoring engines that "imagines" molecules from scratch using cutting-edge deep learning technologies we pioneered for medical use back in 2015. Chemistry42 automatically creates drug-like molecular structures with appropriate physicochemical properties. In this case, Chemistry42 was used to design a library of small molecules that bind to the novel intracellular target revealed by PandaOmics.

With this proof-of-concept study, we demonstrated how an AI-driven drug discovery process can be substantially accelerated and optimized for cost — the whole route from concept to the preclinical drug candidate took the Insilico Medicine team less than 18 month, and a fraction of the cost of a standard drug discovery project.
Important Milestone in Tackling Kidney Fibrosis
After announcing our first PCC for IPF we started to focus on kidney fibrosis as an additional indication.

Choosing kidney fibrosis as a target indication, we confirmed a nanomolar potency of our compounds in kidney cell-based in vitro studies. Consequently, we carried out a proof-of-concept in vivo study – kidney protection after unilateral ureteral obstruction in mice. Already a simple visual inspection of the affected kidneys showed that our molecules improved kidney pathology associated with fibrosis development. Further analysis confirmed that our compounds affected key fibrotic biomarkers and prevented the development of histologically verified fibrosis in murine kidneys. The compounds also demonstrated a good safety profile in a 14-day repeated mouse dose range-finding (DRF) study.

Based on these encouraging results, a drug candidate INS018_073 has been nominated as a preclinical drug candidate in June 2021 for IND-enabling study and further efficacy studies that will lead to clinical investigations. Currently, we have started the scale-up and process development for the candidate compound. We plan to initiate the IND-enabling studies by September 2021 and have them finished by the third quarter of 2022, to submit for Phase I clinical trials at the end of 2022.

The figure below shows a high level summary of experiments performed and results obtained for INS018_073, leading to preclinical candidate nomination.
Improving Drug Discovery —
One Indication at a Time
Since the launch of its fibrosis drug discovery program ISM001, InsilicoMedicine managed to deliver two preclinical candidates for Idiopathic Pulmonary Fibrosis and Kidney Fibrosis within just about 24 months at a cost of around $2 million. The kidney-related subproject was completed within just 6 month with total expenditures of about $300.000.

These results illustrate how integrated AI-driven drug design is a more efficient research paradigm, compared to traditional "big pharma" approach of delivering drug candidates. We managed to dramatically improve both speed and cost of drug discovery, compared to standard estimates for such projects from target selection to the launch of IND-enabling studies.
Stay tuned, follow us on social media!