Ai-powered innovation in drug discovery

The traditional drug discovery process of identifying the causal mechanism of a disease, a single molecular target, and a drug against the target is slow, expensive, and fraught with the

fallout from failed hypotheses. Transcripta Bio, a biotechnology company based in Palo Alto, California, is harnessing the power of artificial intelligence (AI) to sift through its vast

dataset and reduce the risk of failure in drug discovery. Transcripta Bio’s objective is to revolutionize the drug discovery process by screening thousands of compounds across numerous

root-cause genes simultaneously, drastically reducing both time and cost. By leveraging cutting-edge technology, Transcripta Bio aims to create a diverse portfolio of potential therapeutics

for hundreds of diseases. The company is partnering with biopharma and tech companies, research centers and patient foundations seeking innovative and efficient drug discovery solutions. “We

focus on small molecules and the changes they elicit on all the gene readouts present in a cell, known as the transcriptome,” explained Chris Moxham, co-founder, CEO and Chief Scientific

Officer of Transcripta Bio. “Ultimately, most drugs work by changing gene expression in a way that elicits a therapeutic effect; interrogating the transcriptome is a great way to understand

drug response and discover new drugs.” Over the past two years, Transcripta Bio has built a unique database with the results from more than 200 million experiments. By collating the effects

of US Food and Drug Administration (FDA)-approved drugs and small molecules that have reached phase 2/3 clinical trials on gene expression in induced pluripotent stem cell (iPSC)-derived

glutamatergic neurons, the company has generated a Drug–Gene Atlas. “We have mapped the molecular structure of thousands of compounds and their effects on 20,000 genes,” Moxham said. The

Drug–Gene Atlas is not only being used to uncover new therapeutic candidates but also to train machine learning models and carry out virtual screens to find new structures that modulate the

transcriptome for therapeutic benefit. “Our machine learning system has learned the chemistry rules to modulate gene expression, charting a faster and cheaper path in drug discovery,” Moxham

explained. PLATFORM VALIDATION Transcripta Bio has already uncovered five repurposing opportunities for monogenic rare diseases. “Working with patient foundations and physicians, these

drugs have been used off label with encouraging early results,” Moxham said. The company has filed new use patents that cover the use of drugs approved for other indications in patients with

Leigh’s disease, SYNGAP1-related intellectual disability, or autism spectrum disorder caused by a rare gene mutation. “We are eager to expedite phase 2b trials to ensure more patients can

access these treatments,” he added. The company is also screening small molecules that have reached phase 2/3 trials and proven to be safe in humans. By bioinformatically intersecting the

results with the Orphanet database, Transcripta Bio has identified more than 50 compounds that change the expression of genes associated with monogenic disorders in glutamatergic neurons.

“We are now screening the candidates on other cell types, confirming their effects on protein levels and carrying out additional preclinical work in animal or patient-derived preclinical

models of disease to progress them to the clinic,” Moxham said. VIRTUALLY INTERROGATING GENE EXPRESSION TO FUEL NEW DRUG DISCOVERY Transcripta Bio’s Drug–Gene Atlas has been used to train

16,000 machine learning models on the effects of molecules on the expression of 16,000 genes and carry out virtual screens at unprecedented speed and scale. “We completed a virtual screen of

6.5 billion drug-like compounds in a couple of days,” said Clayton Mellina, Chief AI Officer and CTO. “Our Conductor AI modeling suite predicts which structures can tune individual genes

either up or down, while minimizing off-target effects.” When Transcripta Bio synthesized 21 compounds that were predicted to reduce the expression of a particular gene, it found that 20% of

them were active in cells. “We have effectively generated a closed loop system (Fig. 1) that can identify novel portfolios around targets for which industry has high conviction,” Moxham

added. Mellina is excited about the possibilities that the company’s approach affords. “We have reached a stage where we can realize the promise of high-throughput sequencing technologies

for ‘omics’-first drug discovery, and not just for finding drugs to treat monogenic diseases but also reverse signatures of complex polygenic diseases and predict potential side effects,” he

said. Transcripta Bio is interested in engaging with companies seeking to use its pan-therapeutic platform to discover new treatments at ten times capital efficiency and with higher

certainty than traditional drug discovery approaches.