{"id":735,"date":"2025-05-22T13:00:00","date_gmt":"2025-05-22T13:00:00","guid":{"rendered":"https:\/\/www.valencelabs.com\/?p=735"},"modified":"2025-05-22T13:24:15","modified_gmt":"2025-05-22T13:24:15","slug":"txpert-predicting-cellular-responses-to-unseen-genetic-perturbations","status":"publish","type":"post","link":"https:\/\/www.valencelabs.com\/fr\/txpert-predicting-cellular-responses-to-unseen-genetic-perturbations\/","title":{"rendered":"TxPert: Predicting Cellular Responses to Unseen Genetic Perturbations"},"content":{"rendered":"
\n
\n
\n

Introduction<\/h2>\n <\/div>\n
\n
\n

Understanding how cells respond to genetic perturbations is essential to designing effective therapies and accelerating drug development. Yet the traditional approach, screening large numbers of perturbations experimentally across extensive biological contexts (such as different cell types), is expensive and, for all combinations of interest, simply intractable. This often involves a degree of “blind optimization,” where interventions are tested somewhat speculatively without a deep understanding of the underlying cellular machinery, contributing to high failure rates for candidates in late stages of the drug-discovery pipeline.<\/span><\/p>\n

This bottleneck highlights the critical need for ML models capable of learning the underlying rules of cellular behaviour. At Valence Labs, Recursion’s<\/a> AI research engine<\/span>,<\/b> we believe progress hinges on developing models that can accurately predict cellular responses, especially for <\/span>novel<\/span><\/i> interventions or in <\/span>new<\/span><\/i> biological contexts. Out-of-distribution (OOD) generalization to new perturbations and cellular contexts is an essential part of the <\/span>Pr\u00e9diction<\/b> pillar of an effective virtual cell\u2014a vision focused on building models that predict, explain, and help discover biological phenomena. <\/span>
<\/span>
<\/span>See our recent perspective paper<\/a> for more on our vision.\u00a0<\/span><\/p>\n

As an important step in addressing this prediction challenge, we introduce <\/span>TxPert<\/b><\/a>: a state-of-the-art model that leverages multiple biological knowledge networks to accurately predict transcriptional responses under OOD scenarios.<\/span><\/p> <\/div>\n <\/div>\n <\/div>\n<\/div>\n\n\n

\n
\n
\n

TxPert: A Unifying Framework for Transcriptomic Perturbation Prediction<\/h2>\n <\/div>\n
\n
\n

TxPert moves beyond models trained on single datasets, offering a unifying approach that:<\/span><\/p>\n

    \n
  1. \u00a0<\/li>\n<\/ol>\n
      \n
    1. \n
        \n
      1. Learns from diverse datasets:<\/b> It is trained on a broad collection of single-cell perturbation datasets, spanning various experimental techniques and cellular systems.<\/span>

        <\/span><\/li>\n
      2. Excels at multiple OOD Tasks:<\/b> TxPert is specifically engineered to generalize across:

        <\/span>\n