Scientific insights often reflect a shift in perspective. For Envisagenics and its partners, insights may be gained through a more encompassing view of gene expression errors. This view includes errors associated with genetic mutations, as well as errors associated with alternative RNA splicing.
“The challenge was so complex and data-intensive that for years nobody looked into it,” Maria Luisa Pineda, PhD, co-founder and CEO of Envisagenics, explains. Six years ago, when the company was founded, artificial intelligence was in its infancy and machine learning was a little more than a research interest. RNA sequencing (RNA-seq) data was undervalued and underutilized, often limited to confirming discoveries at the level of genes.
Pineda, however, saw potential in artificial intelligence and RNA-seq data. With Martin Akerman, PhD, Envisagenics’ co-founder and CTO, she developed the SpliceCore® machine learning analytics software platform to identify RNA splicing errors. The technology is used for target discovery and identifying potential therapeutics based on splicing errors.
The platform can comb through 1,000 data sets in about two hours, “identifying splicing changes that occur only in certain patient populations,” Pineda informs. Consequently, researchers can correlate splicing errors to diseases and determine their specificity.
Currently, there are 370 known genetic disorders caused by RNA splicing errors. Lung cancer is a case in point. “Today, lung cancer is often diagnosed late,” Pineda notes. “It is driven by splicing errors that can be detected even before the tumor is formed.”
In April, Envisagenics announced a partnership with the Johnson & Johnson Lung Cancer Initiative to build predictive models for lung cancer progression and risk. Within a few months, the partnership expanded Envisagenics’ oncology portfolio and validated the commercial readiness of the SpliceCore technology.
Getting started, says Pineda, is as straightforward as “identifying the scientific interest, determining pain points, and understanding the availability of RNA-seq data.”
“In cancer, half of the synonymous mutations are due to RNA splicing errors,” she continues. “So, we would discuss how the platform works and the problems you may have faced. Oftentimes, you may be targeting the wrong isoform or protein. That will be predicted and identified by SpliceCore.”
“Envisagenics,” Pineda adds, “views exons as independent units of data to enhance sensitivity and scalability when analyzing RNA-seq data.” She points out that from a user’s perspective, the first step is to “map the incoming RNA-seq data to Envisagenics’ exon-centric reference transcriptome.”
This is a computationally intense step that is performed in the cloud. By relying on the cloud, researchers are able to extract value even when working from home, which has been critical during the COVID-19 pandemic.
“We built a solution to perform this step remotely on partners’ clouds in a compliant and secure manner without compromising patient data privacy,” she emphasizes. “After this step, the data is transferred to Envisagenics’ cloud to identify the biological relevance of potential targets by combining different machine learning algorithms. We then prioritize a group of candidates to validate in our wet lab.”
Envisagenics accesses several rich sources of sequencing data: its own two-petabyte in-house repository of cell lines; primary tissues and organoids; public databases; and the databases of its academic and pharmaceutical partners. “The data sets are the enablers,” Pineda advises. “We believe the more data sets you incorporate, the better the results for patients.”
The company is focused on target discovery and compound design for antisense drugs, neoantigen discovery for immunotherapy, and biomarker discovery for patient stratification. The technology itself is both modality- and disease-agnostic. SpliceCore hosts six proprietary algorithms to explore additional areas in oncological, neurodegenerative, and genetic diseases.
Because the machine learning algorithms are trained using proven experimental data, the black box conundrum is eliminated. “We use intuitive features and feature selection techniques to explain how the predictions are made,” Pineda says. “It gives the scientist input,” such as the ability to weight certain features more than others.
Envisagenics was founded in 2014, as cloud computing was becoming commonplace, gene sequencing was becoming practical, and precision medicine was becoming possible. Preliminary results from clinical trials of Spinraza, an antisense oligonucleotide drug for spinal muscular atrophy, were promising. When the science coalesced with these technologies, Pineda concluded the timing was right for a spinout.
Reflecting on how her work evolved after her time as a researcher at Cold Spring Harbor Laboratory, she says that one of the greatest challenges was translating the technology from academic to commercial applications: “You ask different questions in industry than in academia, so you need to consider additional factors such as regulatory compliance and cost effectiveness.”
“The merger of sequencing and RNA therapeutics was brand new at the time we started,” she recalls. “We saw we could automate and accomplish in eight months what (until then) took 12 years.” But being a technological pioneer is not easy. Attracting funding often is challenging—especially for scientific founders without significant business experience.
Envisagenics’ fortunes improved when the company received a grant from the National Institutes of Health. Pineda knew Envisagenics had turned a corner when it received backing from top-tier investors, including Microsoft Ventures and Madrona Venture Group. “Getting those names involved in the company was very important,” Pineda maintains. “They believed in innovating drug discovery methodologies.”
Along the way, Envisagenics won the Johnson & Johnson Innovation–JLABS Artificial Intelligence for Drug Discovery QuickFire Challenge and became the North American winner of Microsoft Ventures’ Innovate.AI competition.
Envisagenics is currently weighing partnership opportunities. “We’re looking for more partnerships with biopharmaceutical companies,” Pineda details. “We have a horizontal platform that can explore multiple disease indications.” Pineda envisions the company facilitating both target and therapeutic discovery with partners to take drug candidates through the clinic.
Envisagenics also plans to advance its in-house antisense oligonucleotide therapeutic program, ENV-0205, for triple-negative breast cancer. Envisagenics has identified a splicing error that occurs in the most aggressive types of breast cancer, in patients with poor prognosis. “We went from an idea to data analysis to compound in eight months,” Pineda declares. “We expect to take the compound into the clinic within the next two years.”
Location: 101 Avenue of the Americas, 3rd Floor, New York, NY 10013
Principal: Maria Luisa Pineda, PhD, Co-Founder and CEO
Number of Employees: 14
Focus: The SpliceCore® software platform combines a proprietary, exon-centric approach to RNA sequencing analysis with machine learning to accelerate the identification and development of therapeutics.