Novel druggable targets for brain conditions

The WHO estimates that 1 in 6 people suffer from brain disorders globally. This translates to over 1 billion people in the world living with disabilities caused by brain conditions and an estimated combined expenditure of over 1 trillion dollars by 2030 in the US alone on disability management in these patient populations. Most neurological conditions currently lack approved drugs, leaving patients with no available treatment options. Therefore, there is a high unmet need for better and faster drug development processes specifically in the space of central nervous system disorders (CNS).

A key step in drug development is the initial drug target discovery stage. The current target discovery approaches in the CNS space have led to notoriously high failure rates in clinical trials in patients. Hence, there is a critical need for the identification of new CNS targets and faster validation of their appropriateness as drug targets to treat or cure CNS conditions.

With the advent of stem cell technologies, omics approaches, and sophisticated data analytics pipelines, we now have robust methods to hunt for new drug targets starting with patient biology (as opposed to animal models that translate poorly to humans). We have the ability to generate vast amounts of data and use AI to help find new targets and better drugs. However, a key challenge in applying powerful AI methods to biological data, is the relatively low numbers of observations/measurements that is usually inherent to most biological data collection experiments. With the exception of a few experimental approaches, most experiments of value are still too costly and too time consuming to be carried out at scale. This is particularly true for data collection from patient specimens that are limited to begin with, and proteomics approaches, which powerfully yield direct insights into the sum of multiple important regulatory layers within the cell. Therefore, we need ways to overcome the problem of limited datasets in biological experiments, to be able to truly take advantage of powerful deep learning methods and apply this to unearth novel druggable targets for brain conditions to help patients with no treatment options

Our solution is a proprietary platform that decodes a novel layer of human CNS biology that we believe is key to most CNS disorders based on our own research and that of others. Our platform combines the most advanced human models of patient biology and a novel omics + AI approach. It reduces noise, enables multi-target/signature identification, and matches signatures back to patients. The novelty of the AI approach lies in overcoming the challenge of using machine learning in biology, with sparse data matrices, such as data acquired from proteomics experiments from a limited set of patients. Briefly, our AI technology allows us to synthetically expand our datasets using Monte-Carlo simulation techniques that take into account distribution of key variables obtained from the dataset. We generate our own in-house high quality datasets. These datasets are then enhanced through a simulation phase to create a machine-learning-ready sample. The sample is then trained with a deep learning algorithm and fed into a jackknifing procedure that allows us to identify defined protein signatures with high probability of disease association. 

Acquisition of large meaningful datasets is not always possible in biology. Our solution allows us to bypass this problem, instead of the more commonly used, yet expensive approach of generating and collecting vast amounts of experimental data. We believe in the power of carefully collected meaningful data, rather than abundant non-focused data and our solution enables the use of such datasets with deep learning methods.

By applying the computational pipeline that we are developing, we can find patterns of protein signatures that best describe the patient across a number of CNS disorders, allowing us to identify novel drug targets, biomarkers, and to better stratify patients for a personalized medicine approach. Our pipeline can eventually be used to predict disease pathogenicity in the absence of a genetic diagnosis and on any type of dataset that is limited by nature. We are first testing and applying this to single gene forms of (rare) neurodevelopmental disorders, because they are genetically defined, with organized and motivated patient populations, and no approved drugs on the market. Thus, our solution helps advance rare patient communities towards the identification of drug targets for their specific conditions and has the potential to yield first-in-class drugs for a number of CNS conditions.

Our solution is initially focused on serving patients with rare neurodevelopmental disorders, but will eventually be applied to larger CNS patient populations. The former are primarily children with severe disabilities including motor, cognitive, and language impairments. The great majority are unable to independently walk, feed themselves, or communicate with their loved ones. Our platform technology will accelerate drug development for these conditions that currently have no treatment options, radically improving the chances of finding precision drugs and in doing so, improving the quality of life of these children and their families. Most pharma companies do not go after these conditions, making it even more important to do this work to help the patients.

Our team consists of myself, a developmental neurobiologist and stem cell scientist with deep subject matter expertise in neurodevelopmental disorders (20+ years). I have for the past 6 years worked with and advised a number of rare disease patient organizations that are driving research in the rare disease space. This has given me a unique view into the patient journey, the challenges of driving research in this space, and close relationships with parents of children with these conditions who provide us with easy access to patient materials, clinical data, and a wealth of unpublished knowledge. This combined with my scientific expertise gives us a competitive advantage. My team also consists of a a machine learning and data analytics expert who is leading the data analytics pipeline and applying existing methods from the field of information systems to the analysis of biological data, a world-renowned expert and Professor of Neurobiology from the Rockefeller University, a founder of a biotech company which was acquired in 2021, with first-hand experience in building from the ground up a successful biotech company, an ex-google product manager with expertise in software and database development and management, and a former managing director from Citi group who is doing the business development. Moreover, we frequently engage with parents and founders of rare disease organizations to get input into our design and thinking of our discovery platform. Together, we have the necessary expertise and drive to build this platform and identify and develop better drugs for the patients. 

Currently, there are no approved drugs to treat the majority of patients that suffer from brain disorders. Therefore there is a high unmet need for novel drug discovery approaches. Our solution is unique in combining state-of-the-art experimental approaches to generate high-quality in-house patient datasets, with a new way to apply AI to decode a novel layer of neurobiology in smaller biological datasets that is characteristic of patient datasets. The AI approach that we are developing could be adapted and applied to other datasets of smaller size and help expand the use of AI in biology in new ways. 

The primary focus of our solution is to identify better drugs, faster, for patients who currently have no treatment options. This aligns with the UN Sustainable Development Goal 3 for good health and well-being for all, by helping to improve the quality of life of patients and their families. Moreover, our AI solution to enable analysis of smaller datasets has the potential to expand access to AI technology for many more populations who are restricted by data sample size because of economic, societal, and biological reasons. 

We are leveraging 1) existing published proteomics datasets 2) our own in-house generated high quality proteomics datasets as we test and refine our deep learning solution for limited datasets. 

The published data are from animal models that have similar mutations as those found in patients with rare neurodevelopmental disorders. The advantage of using these datasets to refine our models is that they are already experimentally validated and therefore serve as an excellent starting point. They are however not as granular as our own in-house datasets that we are generating from patient cells while developing the computational pipeline. We use the published data as preliminary step (Phase 0) to develop and fine-tune our deep learning algorithm and find the correct algorithm parameters (hidden layers, etc.) and the correct jackknifing criteria– i.e., how to create protein signature subsets that are tried in the training phase (see Phase 3, below).

We have deep expertise in human proteomic data generation and have validated pipelines in place to acquire well-controlled data from patients and non-affected controls with multiple technical and biological replicates (Phase 1). Then, the human proteomics data are enhanced by a multitude of Monte-Carlo-simulation-cloned data points (Phase 2). The data enhancement process ensures that we can use cutting edge deep learning approaches, which usually require large sample sizes. 

Particularly, we use the enhanced dataset to train and test over 2,000 deep learning models (Phase 3). Each of these models is created using a subset of protein signatures available in our enhanced datasets, very similar to a jackknifing procedure. We cross-compare these models to identify the critical protein signatures.

We are building a multi-pronged platform technology to radically accelerate drug development for CNS disorders, using the most advanced human models of patient biology and a novel omics + AI approach.

Ensuring ethical and responsible use of AI is a fundamental commitment that guides our research, process development, and operational practices. We maintain transparency in our practices. Working with patient samples and clinical information, we safeguard data privacy and security by adhering to established regulations and industry best practices. We ensure that everyone working with human data has undergone human subject training, which clearly outlines how human data should be treated and de-identified. We are aware of biases in AI algorithms and aim to generate diverse datasets and regularly assess and adjust our models to reduce biases that could impact our analyses. We plan to implement validation and testing steps set out by the NIST to ensure the reliability and accuracy of the AI output. We closely follow and adhere to relevant regulations and guidelines governing AI and healthcare, such as FDA regulations and guidelines set out by the Department of Homeland Security, to ensure that we understand and mitigate potential risks associated with AI in drug discovery. We will in the future establish an Ethics Review Board consisting of experts in ethics, neuroscience, and AI and will rely on this board to provide further guidance on ethical considerations and ensure that our practices align with ethical principles. We actively engage with stakeholders, including patients, clinicians, and advocacy groups, to incorporate their perspectives and feedback into our process development and data handling. By following these practices, we aim to ensure that our use of AI in drug discovery is scientifically sound and ethical, responsible, and aligned with the best interests of patients, healthcare providers, and society.

Our solution has the potential to transform drug development for CNS conditions and massively accelerate efforts at identifying personalized drugs for children with neurodevelopmental disorders in the next five years. 

Our impact goals include:

1. Refining and testing our AI pipeline (year 1) on rare single gene forms of neurodevelopmental disorders as part of our platform development to identify novel drug targets and biomarkers

2. Scaling data generation and the use of our platform technology to at least 10 new single gene syndromes/year (year 2-3)

3. Identifying and validating novel drug targets for drug development, aiming for 1 drug to reach phase I clinical trials by year 5

4. Applying our platform to CNS conditions with complex genetics (years  4-5)

In the first year, we will have made significant progress on developing and testing our novel AI approach that takes advantage of the power of AI but enables the analysis of smaller datasets, characteristic of many biological datasets for rare diseases (and stratified patients). Our goal is to extract critical insights into patient biology, from in-house generated high quality datasets,  to enable novel drug target discovery for patients. Rare (genetic) neurodevelopmental disorders are typically neglected populations by the pharma industry and traditional drug developers as larger populations are financially more lucrative. There are currently 800 rare neurodevelopmental disorders with no known cure or treatment options. By developing a platform that can be applied across many of these conditions we are tackling a neglected space with high unmet need for patients. One of the defining features of these conditions is the significant symptom overlap across distinct genetic conditions. As we expand and scale the use of our platform technology (in years 2-3) across more neurodevelopmental disorders that all share commonalities in symptom presentation, our platform has the potential to identify targets that work across populations on specific symptoms, hence paving the way for re-grouping patients and developing personalized medicine in a new way. While our initial focus population are patients with rare (genetic) neurodevelopmental disorders, once validated, we plan to apply the platform to idiopathic cases, which account for the great majority of CNS patients. For example, there are currently 50 million children with autism in the world, 85% of which are idiopathic cases. We are building a solution that has the potential to tackle genetically complex disorders and therefore transform CNS drug development. 

1 full time scientist

1 part time scientist (remote)

1 part time machine learning/data scientist lead

1 part time data scientist student (paid)

1 part time business consultant

1 part time operations consultant

(Core members of the team are currently not drawing salaries)

We have been working on our solution for over a year, with the first person to go full time, 6 months ago

Being a female founder of immigrant background, I have a wide perspective of the challenges faced by minorities, but also have firsthand lived experience of how diversity in people can help open up new thinking and perspectives in science and in business. We are committed to building a company with a diverse workforce, equal compensation, and equity and inclusivity in engagement. We cultivate a culture where differences are celebrated and encouraged. Our team already consists of members with different nationalities, ethnicities, genders, and orientations and we plan to keep growing the team in that way. In addition, we strongly believe in opening doors and giving opportunities to underrepresented minority trainees (next generation) and we will do so by taking in interns from the LifeSci NYC internship program. 

We are a nimble organization, fully focused on drug target identification and development for neurodevelopmental and psychiatric conditions through a novel discovery platform. The founder has deep subject matter expertise and technical knowhow in the fields of neurobiology, stem cell biology and protein biochemistry, which are foundational to our core technology and in-house data generation component of the platform. Further, we have on-boarded world-renowned scientific and data science advisors and are planning to carry out a sponsored research agreement with the Rockefeller University to leverage key insights to optimize our core technology. The team also consists of an academic collaborator from McGill University who will continue to work with us on developing and refining our novel AI solution and we plan to bring in additional expert advisors in this space. We have also on-boarded three experienced business advisors who have startup creation experience to help us navigate the early steps of company creation and partnership developments. We have established close relationships with a number of patient organizations whom we serve, and keeping an active dialogue with them allows us to gain deeper understanding of patient needs and access to vital patient material, clinical information, and unpublished knowledge. In return, we provide advice on what needs to be built to further research into their respective conditions and how to best get clinical-trial ready. We plan to establish further relationships with pharma executives through warm introductions and conference attendances, to help guide our platform development and pave the way for future potential partnerships.       

We plan to initially cover expenses through various government grants (SBIRs) and Angel investors as well as small-scale fee-for-service projects to put us in a position to raise a seed round. We have traction with patient organizations for fee-for-service projects. In the next phase, we plan to operate on a licensing model to pharma, where we provide novel validated CNS targets and biomarkers. In the past year, there have been two examples of AI-powered target discovery platforms that have struck >$100M licensing deals with pharma specifically for novel CNS targets. Our long-term plan is to develop our own in-house CNS drug pipeline to bring to market for patients in partnership with patient organizations. 

We are currently bootstrapping with minimal operating costs and are generating small amounts of revenue to continue building our data analytics platform through discounted fee-for-service projects ($15K so far). If we obtain funding through this challenge, we will continue to bootstrap and use the funds to develop our AI solution, while we actively fundraise from both dilutive and non-dilutive sources to start our wet lab operation for in-house data generation. At that stage we are planning to operate on a projected budget of $1.2M (includes a ~20% margin) per year.

In the first year, that would allow us to fully on-board and pay:

2 full-time scientists ($168K + $144K in salaries and benefits)

1 full-time ML expert/data analyst ($144K in salary and benefits)

Generate in-house datasets from patient samples (23 specimens) collected for 1 syndrome and test platform end-to-end ($360K) 

IP and legal costs ($35K)

Wet lab space ($108K)

Cloud services ($40K)

We are requesting $100K to allow us to continue to develop and iterate our AI solution as part of our CNS target discovery platform

-$40K costs for cloud services and data management

-$10K salary cost for a contracted machine learning graduate student to run analyses

-$50K to generate an in-house high quality dataset for further testing and iteration of the AI solution.

These costs are estimated from obtained quotes. 

A Cure residency would significantly propel and advance the development of our core platform technology, at our current stage. In particular, we would benefit immensely from access to the Cure drug development infrastructure and expertise. As we plan to generate our own in-house datasets, access to lab space and seed funding would be an invaluable benefit that would provide much needed support and put us in a position to raise the next round of capital. The networking and mentorship opportunities would also be of immense help at our stage as we strive to establish further partnerships with drug developers. As an NYC-based life sciences startup with deep connections to established academic institutions (the Rockefeller University) and relationships with NYC-residing talent whom we would love to work with, we cannot think of a better environment in New York City than the Cure to help us launch this mission-driven venture.