AI-powered genomics tool, SCADE, significantly outperforms by more than doubling the accuracy of generic models, including those from OpenAI and DeepSeek, in solving single-cell genomics problems
Valted Seq, Inc., a pioneer in cutting-edge biotechnology solutions, today announced the launch of a novel AI-powered genomics tool, Single Cell AI Discovery Engine (SCADE), designed to revolutionize how researchers and healthcare professionals interpret vast amounts of genetic data. By leveraging advanced machine learning algorithms, SCADE promises to drastically improve the speed and accuracy of insights derived from genomic sequences paving the way for faster biomarker and drug discovery, personalized medicine, and more effective treatments for cellular diseases.
“Our goal is to enable scientists such as biologists and biochemists to perform genomics and bioinformatics without writing computer code,” said Bardia Nezami, CEO of Valted Seq. “Analyzing genomics data, let alone from single cells, is incredibly complex. However, with SCADE we are democratizing that process. It enables robust analysis and interpretation at a scale never before possible expediting the journey toward life-saving breakthroughs. We have made this possible by training our AI models on high-quality homogenous single cell big data generated in-house consisting of tens of millions of cells.”
Generic AI models (OpenAI, DeepSeek, Claude, Gemini, etc) fail to solve complex genomics problems because of a lack of specific domain knowledge and/or fine-tuning. Valted Seq aims to address this issue by training specialized AI engines for different genomics tasks, applications, and tissue types by utilizing highly homogeneous in-house-generated single-cell big data. SCADE, with hierarchical multi-layer models consisting of data management, bioinformatics core, basic large language model (LLM), specialized genomics layer, specialized single cell layer, and application-dependent layer, offers custom-tailored AI tools for specific use cases such as neurology, oncology, microbial, or immune system.
One of Valted Seq’s key contributions is the creation of a benchmark database specifically designed for evaluating AI models on single cell analysis tasks. This is accomplished by focusing on foundational categories, such as cell type annotation, which includes hundreds of problems. This benchmark establishes an important baseline for measuring performance in this specialized domain. Valted Seq is expanding the benchmark database to thousands of problems.
Valted Seq’s first phase testing revealed that standard AI approaches struggle significantly with complex tasks, with even the best conventional model achieving only about 25% code execution success and around 14% accuracy in matching expected results. In contrast, Valted Seq’s SCADE framework, using a multi-agent framework and trained on tens of millions of single cells generated in-house, can dramatically improve the results beyond 96% code execution success and 36% accuracy in matching expected results. SCADE more than doubles the accuracy of standalone models.
Key features of SCADE include:
- Genomics Data at Scale: Generates high-quality big data for training and fine-tuning AI genomics and bioinformatics tools via robotics.
- High-Speed Data Processing: Deploys advanced AI algorithms to simultaneously analyze multiple large-scale datasets, significantly reducing the time it takes to identify variations and patterns.
- Unprecedented Accuracy: Employs state-of-the-art models that minimize false positives and negatives, boosting confidence in critical decisions.
- User-Friendly Dashboard: Equips scientists with an intuitive interface for exploring and analyzing genomic data, as well as creating reports unprompted, helping teams rapidly uncover hidden relationships and reliable targets.
- Scalability for Research and Clinical Settings: Able to handle both small-scale lab studies and large-scale projects, ensuring accessibility and flexibility for a wide range of users.
By offering unparalleled speed and accuracy, SCADE provides a powerful advantage for organizations seeking to stay at the forefront of precision medicine. Pharmaceutical and biotechnology companies can utilize the platform to shorten the time from drug discovery to market while healthcare providers can refine treatment strategies based on individual genomic profiles.
“Scientists need clear, actionable insights, and they need them quickly,” added Bardia Nezami. “AI-driven platforms like SCADE are no longer just ‘nice to have’; they are rapidly becoming the foundation upon which modern genomic science stands. Researchers worldwide are seeking innovative tools that can streamline the interpretation of complex genomics data at very large scale and help them keep pace with rapid advancements in precision medicine.”
For more information, please contact info@valtedseq.com.
Background
Single cell analysis, the frontier of molecular biology, is commercially dominated by droplet-based technology. However, the state-of-the-art droplet-based approach comes with two major problems regarding accuracy and scalability due to limits of the underlying physics. Although the alternative combinatorial approach can address the accuracy problem to a good degree and the scalability problem to some extent, it still lacks the automation which is an integrated feature of the droplet-based method already available in the market.
On the other hand, the lack of big data level analysis tools is an evident bottleneck and opportunity for commercial applications in this area which is a direct result of limited or low scalability of the available products from existing methods. Like other fields, without big data, making AI-powered single cell discovery tools is quite challenging, if not out of reach. Valted Seq is in a unique position to address these challenges by developing a robotic-based automation platform for sample processing which generates high-quality large-scale homogenous data required to train AI-powered analysis tools for single cell genomics.
Valted Seq has managed to build one of the largest databases of tens of millions of single cells and aims to expand to billions to improve the performance of AI models. This data is generated from post-mortem frozen brain samples acquired from Johns Hopkins University and Banner Health Institute biobanks, the top two internationally recognized brain repositories. Valted Seq has already identified several novel cell type-specific biomolecules for different diseases as biomarkers and therapeutic targets using this database. Most have already been verified in vitro and in vivo in a preclinical setting.