Business description of Kiromic-Biopharma-Inc from last 10-k form
ITEM 1. BUSINESS.
Overview
Revolutionizing Next-Gen Allogenic CAR Therapies for Solid Tumors.
We are a target discovery and gene-editing company utilizing artificial intelligence and our proprietary neural network platform with a therapeutic focus on immuno-oncology.
Our proprietary target discovery engine is called "Diamond."
Kiromic's Diamond is big data science meeting target identification, dramatically compressing man-years and billions of drug development dollars to develop a live drug.
Without Kiromic's Diamond, the management of all the data required to solve the Target Identification puzzle is both challenging and inefficient. Normal data required for target identification would require manual analysis of thousands of cancer tissue samples with billions of data points, looking at millions of mutations, and poring over thousands of publications on oncology and targets.
Diamond (Screening, Prioritizing, and Harmonizing)
Diamond is a computational platform and a neural network that can identify new cancer immunological targets for T cells and B cells. Diamond is an artificial intelligence and machine learning approach that can identify novel surface tumor targets. It uses public and proprietary samples and can expand into the tumor target space.
Diamond addresses the main challenges in today's clinical pipeline: target identification.
Diamond generates a prioritized list of cancer immunological targets for T cells and B cells. These targets can be used to create therapies such as antibody therapies, T cell therapies, T cell receptor therapies, CAR-T cell therapies and vaccine therapies.
Diamond's cognitive and deep learning capabilities extract information from our extensive digital library consisting of clinical studies, genomic and proteomic datasets. Diamond harmonizes all the raw data and creates datasets, which allows us to screen for cancer targets. Diamond will identify and prioritize lists of genes (biomarkers, wild type, mutant, isoform, neoepitope, etc.) that are highly and specifically expressed in the disease of interest while providing its
3
distribution and methylation status across the entire patient population. It also maps out the exact portion of the gene that will elicit an immune response.
Diamond performs meta-analysis and convolution studies while standardizing and normalizing data across multiple and variable experimental platforms, then allows for the visualization of consistent and accurate results in a user-friendly fashion.
See our Diagram below which will walk readers through our process of going from antigens and target libraries to finish with target selection by our artificial intelligence engine.
4
5
Prioritizing T and B Cell Targets. Diamond generates a prioritized list of immunological targets for T cells and B cells. These targets can be used to create therapies such as antibody therapies, T cell therapies, T cell receptor therapies, CAR T cell therapies and vaccine therapies.
Identify Highly Expressed Genes. Diamond's cognitive and deep learning capabilities extract information from our extensive digital library consisting of clinical studies, genomic and proteomic datasets. Diamond harmonizes all the raw data and creates datasets which allows us to screen for cancer targets. Diamond will identify and prioritize lists of genes (biomarkers, wild type, mutant, isoform, neoepitope, etc.) that are highly and specifically expressed in the disease of interest while providing its distribution status across the entire patient population. It also maps out the exact portion of the gene that will elicit an immune response.
Performs Meta Analysis. Diamond performs meta-analysis and convolution studies while standardizing and normalizing data across multiple and variable experimental platforms, then allows for the visualization of consistent and accurate results in a user-friendly fashion.
Predict Isoform Targets. Cancer cells will down regulate or shed targets in order to avoid detection and destruction by T cells (the immune system). These variations are known as isoforms. CancerSplice also shows a box plot by tissue of expression of the isoform in normal cancer genome atlas tissues and a box plot of the matching isoform in genotype-tissue expression program normal data. The sequence of amino acids that are specific for the selected cancer isoforms are then directly fed to Diamond's artificial neural capsule network for peptide design and prioritization.
CancerSplice (Isoform Target Prediction)
Cancer cells will down-regulate or shed targets in order to avoid detection and destruction by T cells (the immune system). One mechanism for this tumor defense is the selection for alternative splice forms of target proteins. These variations are known as isoforms. Target isoforms include variations in their primary amino acid sequence that can change both the final folded form of the target plus their ability to be recognized by pre-existing and modified T cells. Within a heterogeneous cancer cell population, isoforms can preferentially expand to avoid detection and destruction by T cells. These isoforms can make it impossible for T cells to outright bind the targets on cancer cells. No binding to the target means no killing of cancer cells.
6
To solve the problem of identifying shared, common cancer-specific antigens derived from alternative splicing and cancer-specific isoform formation, we have developed a fully integrated in silico methodology to predict cancer-specific isoforms called CancerSplice.
CancerSplice allows for the prediction and prioritization of iso-antigens which could serve as a novel source of tumor targets, highly specific for neoplastic cells but without the drawback of also being highly patient-specific.
CancerSplice allows the user to select a tissue type from the cancer genome atlas along with thresholds for filtering isoforms (minimum and maximum tumor and normal cell transcript parts per million). Based on the tissue selected, CancerSplice displays a sorted list of isoforms that are elevated in high-expressing tumors versus normal tissues which have low expression. Differential analysis is then performed and used to generate two types of lists: (1) isoforms expressed in tumor but not expressed in normal tissues; and (2) isoforms expressed in normal tissues but yet at a much higher level in tumors. CancerSplice then allows the user to click on an isoform in the list to select a specific isoform to display in a detailed panel, which shows the multi-sequence alignment for the isoform, as well as all the other isoforms of that gene.
Finally, CancerSplice also shows a box plot by tissue of expression of the isoform in normal cancer genome atlas tissues and a box plot of the matching isoform in genotype-tissue expression program normal data. The sequence of amino acids that are specific for the selected cancer isoforms are then directly fed to Diamond's artificial neural capsule network for peptide design and prioritization.
Therefore, we believe that we have developed unique tools to address the issue with tumor-specific iso-antigens through CancerSplice and Diamond.
7