Utilizing machine learning and digital embedding processes to generate digital maps of biology and user interfaces for evaluating map efficacy
The present disclosure relates to systems, non-transitory computer-readable media, and methods for utilizing machine learning and digital embedding processes to generate digital maps of biology and user interfaces for evaluating map efficacy. In particular, in one or more embodiments, the disclosed systems receive perturbation data for a plurality of perturbation experiment units corresponding to a plurality of perturbation classes. Further, the systems generate, utilizing a machine learning model, a plurality of perturbation experiment unit embeddings from the perturbation data. Additionally, the systems align, utilizing an alignment model, the plurality of perturbation experiment unit embeddings to generate aligned perturbation unit embeddings. Moreover, the systems aggregate the aligned perturbation unit embeddings to generate aggregated embeddings. Furthermore, the systems generate perturbation comparisons utilizing the perturbation-level embeddings.
1. A computer-implemented method comprising:
receiving, from a client device, a request for a benchmark measure for a perturbation embedding model;
identifying perturbation data for the perturbation embedding model comprising a plurality of perturbation unit embeddings generated by the perturbation embedding model from a plurality of perturbation experiment units;
determining, utilizing a benchmark model, a univariate benchmark measure from the plurality of perturbation unit embeddings by utilizing a first perturbation unit embedding from the plurality of perturbation unit embeddings corresponding to a shared perturbation class and a second perturbation unit embedding from the plurality of perturbation unit embeddings corresponding to the shared perturbation class by:
determining a first similarity measure between the first perturbation unit embedding and the second perturbation unit embedding;
determining a second similarity measure between the first perturbation unit embedding and a third perturbation unit embedding corresponding to the shared perturbation class;
determining a third similarity measure between the second perturbation unit embedding and the third perturbation unit embedding; and
determining a combined similarity measure for the shared perturbation class from the first similarity measure, the second similarity measure, and the third similarity measure; and
providing, for display via a user interface of the client device, the univariate benchmark measure for the perturbation embedding model.
2. The computer-implemented method of claim 1 , wherein
providing the univariate benchmark measure comprises providing a consistency metric or a distance metric for the plurality of perturbation unit embeddings.
3. The computer-implemented method of claim 2 , wherein providing the consistency metric comprises:
providing a consistency of observed perturbation effects on cells or subjects within the shared perturbation class; or
providing a signal recovery of the shared perturbation class relative to a non-perturbation control class.
4. The computer-implemented method of claim 2 , wherein utilizing the first perturbation unit embedding and the second perturbation unit embedding comprises at least one of: determining a cosine similarity between the first perturbation unit embedding and the second perturbation unit embedding or determining a projection similarity between the first perturbation unit embedding and the second perturbation unit embedding.
5. The computer-implemented method of claim 1 , further comprising determining, utilizing the benchmark model, a multivariate benchmark measure for the perturbation embedding model comprises comparing a predicted measure of bioactivity between a first perturbation class and a second perturbation class with an observed measure of bioactivity between the first perturbation class and the second perturbation class.
6. The computer-implemented method of claim 5 , wherein identifying the perturbation data for the perturbation embedding model comprises identifying a first perturbation-level aggregated embedding for the first perturbation class and a second perturbation-level aggregated embedding for the second perturbation class.
7. The computer-implemented method of claim 6 , further comprising determining the predicted measure of bioactivity between the first perturbation class and the second perturbation class by comparing the first perturbation-level aggregated embedding for the first perturbation class and the second perturbation-level aggregated embedding for the second perturbation class.
8. The computer-implemented method of claim 5 , further comprising providing, for display via the user interface of the client device, the univariate benchmark measure for the perturbation embedding model and the multivariate benchmark measure for the perturbation embedding model.
9. 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:
receive, from a client device, a request for a benchmark measure for a perturbation embedding model;
identify perturbation data for the perturbation embedding model comprising a plurality of perturbation unit embeddings generated by the perturbation embedding model from a plurality of perturbation experiment units;
determine, utilizing a benchmark model, a multivariate benchmark measure from the plurality of perturbation unit embeddings by comparing a predicted measure of bioactivity between a first perturbation class and a second perturbation class with an observed measure of bioactivity between the first perturbation class and the second perturbation class; and
provide, for display via a user interface of the client device, the multivariate benchmark measure for the perturbation embedding model.
10. The system of claim 9 , further comprising instructions that, when executed by the at least one processor, cause the system to determine, utilizing the benchmark model, a univariate benchmark measure for the perturbation embedding model by comparing a first perturbation unit embedding from the plurality of perturbation unit embeddings corresponding to a shared perturbation class with a second perturbation unit embedding from the plurality of perturbation unit embeddings corresponding to the shared perturbation class.
11. The system of claim 10 , wherein the instructions, when executed by the at least one processor, cause the system to compare the first perturbation unit embedding with the second perturbation unit embedding by at least one of: determining a cosine similarity between the first perturbation unit embedding and the second perturbation unit embedding or determining a projection similarity between the first perturbation unit embedding and the second perturbation unit embedding.
12. The system of claim 9 , wherein the instructions, when executed by the at least one processor, cause the system to determine, utilizing the benchmark model, the multivariate benchmark measure for the perturbation embedding model by comparing a predicted measure of bioactivity between a first perturbation class and a second perturbation class with an observed measure of bioactivity between the first perturbation class and the second perturbation class.
13. The system of claim 12 , wherein the instructions, when executed by the at least one processor, cause the system to identify the perturbation data for the perturbation embedding model by identifying a first perturbation-level aggregated embedding for the first perturbation class and a second perturbation-level aggregated embedding for the second perturbation class.
14. The system of claim 13 , further comprising instructions that, when executed by the at least one processor, cause the system to: determine the predicted measure of bioactivity between the first perturbation class and the second perturbation class by comparing the first perturbation-level aggregated embedding for the first perturbation class and the second perturbation-level aggregated embedding for the second perturbation class.
15. The system of claim 12 , further comprising instructions that, when executed by the at least one processor, cause the system to: provide, for display via the user interface of the client device, a univariate benchmark measure for the perturbation embedding model and the multivariate benchmark measure for the perturbation embedding model.
16. A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause a computing device to:
receive, from a client device, a request for a benchmark measure for a perturbation embedding model;
identify perturbation data for the perturbation embedding model comprising a plurality of perturbation unit embeddings generated by the perturbation embedding model from a plurality of perturbation experiment units;
determine, utilizing a benchmark model, a multivariate benchmark measure from the plurality of perturbation unit embeddings by comparing a predicted measure of bioactivity between a first perturbation class and a second perturbation class with an observed measure of bioactivity between the first perturbation class and the second perturbation class; and
provide, for display via a user interface of the client device, the multivariate benchmark measure for the perturbation embedding model.
17. The non-transitory computer-readable medium of claim 16 , further comprising instructions that, when executed by the at least one processor, cause the computing device to: determine, utilizing the benchmark model, a univariate benchmark measure for the perturbation embedding model by comparing a first perturbation unit embedding from the plurality of perturbation unit embeddings corresponding to a shared perturbation class with a second perturbation unit embedding from the plurality of perturbation unit embeddings corresponding to the shared perturbation class.
18. The non-transitory computer-readable medium of claim 17 , wherein the instructions, when executed by the at least one processor, cause the computing device to compare the first perturbation unit embedding with the second perturbation unit embedding by:
determining a first similarity measure between the first perturbation unit embedding and the second perturbation unit embedding;
determining a second similarity measure between the first perturbation unit embedding and a third perturbation unit embedding corresponding to the shared perturbation class;
determining a third similarity measure between the second perturbation unit embedding and the third perturbation unit embedding; and
determining a combined similarity measure for the shared perturbation class from the first similarity measure, the second similarity measure, and the third similarity measure.
19. The non-transitory computer-readable medium of claim 17 , wherein the instructions, when executed by the at least one processor, cause the computing device to compare the first perturbation unit embedding with the second perturbation unit embedding by at least one of: determining a cosine similarity between the first perturbation unit embedding and the second perturbation unit embedding or determining a projection similarity between the first perturbation unit embedding and the second perturbation unit embedding.
20. The non-transitory computer-readable medium of claim 16 , wherein the instructions, when executed by the at least one processor, cause the computing device to determine, utilizing the benchmark model, the multivariate benchmark measure for the perturbation embedding model by comparing a predicted measure of bioactivity between a first perturbation class and a second perturbation class with an observed measure of bioactivity between the first perturbation class and the second perturbation class.