Utilizing machine learning models to synthesize perturbation data to generate perturbation heatmap graphical user interfaces
The present disclosure relates to systems, non-transitory computer-readable media, and methods for embedding perturbation data via a machine learning model and filtering, aligning, and aggregating the embeddings to generate a genome-wide perturbation database for real-time generation of perturbation heatmaps. In particular, in one or more embodiments, the disclosed systems can receive a plurality of perturbation images portraying cells from a plurality of wells corresponding to a plurality of cell perturbations. Further, the systems can generate, utilizing a machine learning model, a plurality of well-level image embeddings from the plurality of perturbation images. Moreover, the systems can align, utilizing an alignment model, the plurality of well-level image embeddings to generate aligned well-level image embeddings. Additionally, the systems can aggregate, according to perturbations of one or more perturbation experiments, the well-level image embeddings to generate perturbation-level image embeddings. Furthermore, the systems can generate perturbation comparisons utilizing the perturbation-level image embeddings.
1. A computer-implemented method comprising:
identifying a plurality of initial cell image embeddings;
aggregating the plurality of initial cell image embeddings according to cell perturbations to generate cell image embeddings;
generating a dataframe comprising the cell image embeddings corresponding to the cell perturbations;
receiving, from a client device, a similarity query comprising a plurality of cell perturbations;
in response to receiving the similarity query, accessing cell image embeddings corresponding to the plurality of cell perturbations from the dataframe;
determining similarity measures for the plurality of cell perturbations from the cell image embeddings corresponding to the plurality of cell perturbations of the similarity query; and
transmitting, to the client device, a query response comprising the similarity measures from the cell image embeddings.
2. The computer-implemented method of claim 1 , further comprising:
generating an additional dataframe comprising aggregated embedding metadata of the cell image embeddings of the dataframe;
identifying an additional plurality of initial cell image embeddings incorporated into a perturbation database and metadata of the additional plurality of initial cell image embeddings; and
regenerating the dataframe to include the additional plurality of initial cell image embeddings and the additional dataframe to include the metadata of the additional plurality of initial cell image embeddings.
3. The computer-implemented method of claim 1 , further comprising:
aggregating metadata corresponding to the plurality of initial cell image embeddings to generate aggregated embedding metadata;
generating an additional dataframe comprising the aggregated embedding metadata of the cell image embeddings of the dataframe;
accessing the cell image embeddings from the dataframe and corresponding aggregated embedding metadata from the additional dataframe; and
transmitting the query response by transmitting the cell image embeddings and the aggregated embedding metadata to the client device.
4. The computer-implemented method of claim 3 , wherein determining the similarity measures for the plurality of cell perturbations from the cell image embeddings comprises at least one of:
comparing a first cell image embedding of the cell image embeddings with a second cell image embedding of the cell image embeddings; or
comparing the cell image embeddings corresponding to the plurality of cell perturbations of the similarity query with additional cell image embeddings corresponding to additional cell perturbations.
5. The computer-implemented method of claim 4 , further comprising:
receiving additional cell image embeddings; and
aggregating the additional cell image embeddings and the plurality of initial cell image embeddings to generate updated cell image embeddings.
6. The computer-implemented method of claim 5 , further comprising:
receiving, from an additional client device, an additional similarity query comprising an additional set of cell perturbations;
accessing the updated cell image embeddings corresponding to the additional set of cell perturbations; and
determining additional similarity measures for the additional set of cell perturbations from the updated cell image embeddings corresponding to the additional set of cell perturbations of the additional similarity query.
7. The computer-implemented method of claim 1 , wherein determining the similarity measures between the plurality of cell perturbations by comparing the cell image embeddings corresponding to the plurality of cell perturbations of the similarity query comprises determining a cosine similarity between the cell image embeddings corresponding to the plurality of cell perturbations of the similarity query.
8. The computer-implemented method of claim 1 , wherein receiving, from the client device, the similarity query comprising the plurality of cell perturbations comprises receiving at least one of a gene perturbation query or a compound perturbation query.
9. The computer-implemented method of claim 1 , wherein transmitting, to the client device, the query response comprising the similarity measures comprises transmitting, to the client device, a perturbation heatmap portraying the similarity measures.
10. A system comprising:
at least one processor; and
at least one non-transitory computer-readable storage medium storing instructions that, when executed by the at least one processor, cause the system to:
identify a plurality of initial cell image embeddings;
aggregate the plurality of initial cell image embeddings according to cell perturbations to generate cell image embeddings;
generate a dataframe comprising the cell image embeddings corresponding to the cell perturbations;
receive, from a client device, a similarity query comprising a plurality of cell perturbations;
in response to receiving the similarity query, access cell image embeddings corresponding to the plurality of cell perturbations from the dataframe;
determine similarity measures for the plurality of cell perturbations from the cell image embeddings corresponding to the plurality of cell perturbations of the similarity query; and
transmit, to the client device, a query response comprising the similarity measures from the cell image embeddings.
11. The system of claim 10 , further comprising instructions that, when executed by the at least one processor, cause the system to:
generate an additional dataframe comprising aggregated embedding metadata of the cell image embeddings of the dataframe;
identify an additional plurality of initial cell image embeddings incorporated into a perturbation database and metadata of the additional plurality of initial cell image embeddings; and
regenerate the dataframe to include the additional plurality of initial cell image embeddings and the additional dataframe to include the metadata of the additional plurality of initial cell image embeddings.
12. The system of claim 10 , further comprising instructions that, when executed by the at least one processor, cause the system to:
aggregate metadata corresponding to the plurality of initial cell image embeddings to generate aggregated embedding metadata;
generate an additional dataframe comprising the aggregated embedding metadata of the cell image embeddings of the dataframe;
access the cell image embeddings from the dataframe and corresponding aggregated embedding metadata from the additional dataframe; and
transmit the query response by transmitting the cell image embeddings and the aggregated embedding metadata to the client device.
13. The system of claim 12 , further comprising instructions that, when executed by the at least one processor, cause the system to determine the similarity measures for the plurality of cell perturbations from the cell image embeddings by performing at least one of:
comparing a first cell image embedding of the cell image embeddings with a second cell image embedding of the cell image embeddings; or
comparing the cell image embeddings corresponding to the plurality of cell perturbations of the similarity query with additional cell image embeddings corresponding to additional cell perturbations.
14. The system of claim 13 , further comprising instructions that, when executed by the at least one processor, cause the system to:
receive additional cell image embeddings; and
aggregate the additional cell image embeddings and the plurality of initial cell image embeddings to generate updated cell image embeddings.
15. The system of claim 14 , further comprising instructions that, when executed by the at least one processor, cause the system to:
receive, from an additional client device, an additional similarity query comprising an additional set of cell perturbations;
access the updated cell image embeddings corresponding to the additional set of cell perturbations; and
determine additional similarity measures for the additional set of cell perturbations from the updated cell image embeddings corresponding to the additional set of cell perturbations of the additional similarity query.
16. A non-transitory computer-readable storage medium storing instructions that, when executed by at least one processor, cause a computing device to:
identify a plurality of initial cell image embeddings;
aggregate the plurality of initial cell image embeddings according to cell perturbations to generate cell image embeddings;
generate a dataframe comprising the cell image embeddings corresponding to the cell perturbations;
receive, from a client device, a similarity query comprising a plurality of cell perturbations;
in response to receiving the similarity query, access cell image embeddings corresponding to the plurality of cell perturbations from the dataframe;
determine similarity measures for the plurality of cell perturbations from the cell image embeddings corresponding to the plurality of cell perturbations of the similarity query; and
transmit, to the client device, a query response comprising the similarity measures from the cell image embeddings.
17. The non-transitory computer-readable storage medium of claim 16 , further comprising instructions that, when executed by the at least one processor, cause the computing device to:
generate an additional dataframe comprising aggregated embedding metadata of the cell image embeddings of the dataframe;
identify an additional plurality of initial cell image embeddings incorporated into a perturbation database and metadata of the additional plurality of initial cell image embeddings; and
regenerate the dataframe to include the additional plurality of initial cell image embeddings and the additional dataframe to include the metadata of the additional plurality of initial cell image embeddings.
18. The non-transitory computer-readable storage medium of claim 16 , further comprising instructions that, when executed by the at least one processor, cause the computing device to:
aggregate metadata corresponding to the plurality of initial cell image embeddings to generate aggregated embedding metadata;
generate an additional dataframe comprising the aggregated embedding metadata of the cell image embeddings of the dataframe;
accessing the cell image embeddings from the dataframe and corresponding aggregated embedding metadata from the additional dataframe; and
transmitting the query response by transmitting the cell image embeddings and the aggregated embedding metadata to the client device.
19. The non-transitory computer-readable storage medium of claim 18 , further comprising instructions that, when executed by the at least one processor, cause the computing device to:
receive additional cell image embeddings; and
aggregate the additional cell image embeddings and the plurality of initial cell image embeddings to generate updated cell image embeddings.
20. The non-transitory computer-readable storage medium of claim 19 , further comprising instructions that, when executed by the at least one processor, cause the computing device to:
receive, from an additional client device, an additional similarity query comprising an additional set of cell perturbations;
access the updated cell image embeddings corresponding to the additional set of cell perturbations; and
determine additional similarity measures for the additional set of cell perturbations from the updated cell image embeddings corresponding to the additional set of cell perturbations of the additional similarity query.