CellNet is a network biology-based computational platform that more accurately assesses the fidelity of cellular engineering than existing methodologies and generates hypotheses for improving cell derivations.

You can use CellNet online by uploading your data here, or you can download and run CellNet locally. You can also use this site to search for predicted transcriptional targets of over 1,200 mouse or human transcriptional regulators.

We applied CellNet to expression data from 56 published reports. This analysis showed that cells derived via directed differentiation more closely resemble their in vivo counterparts than products of direct conversion, as reflected by the establishment of target cell-type GRNs. We have also used the platform to experimentally demonstrate an unanticipated developmental potential of directly converted hepatocytes, and to improve the function of directly converted macrophages. While we have mainly used CellNet in cell engineering, it will also yield insights into the dysregulation of normal transcriptional programs in pathological states, including cancer.


CellNet takes as input a gene expression profile and returns:

  1. Classification values estimating the likelihood that the profile comes from one of 16 (human) or 20 (mouse) cell- and tissues types. Classification scores are stringent criteria to assess the extent to which an engineered population resembles the training data.
  2. Network status, which indicates the extent to which a cell or tissue type GRN is established in the gene expression profile. The GRN status is a sensitive metric of the extent to which specific GRNs are induced or repressed in different conditions.
  3. Network influence scores for all transcriptional regulators reflecting the extent to which a transcriptional regulator and its target genes are dysregulated in the query sample, weighted by the importance of the regulator the cell and tissue specific GRN. These scores can be used to prioritize candidate factors to iteratively improve cell engineering.