Systems and methods for evaluating whether perturbations discriminate an on target effect

A method for determining whether a set of test chemical compositions discriminates over a null distribution for an on target effect against a first gene of a cell, wherein the set of test chemical compositions comprises a plurality of test chemical compositions against the first gene and the cell comprises a plurality of genes including the first gene, the method comprising:

• at a computer system comprising a processor and a memory:
• (A) obtaining, for each respective test chemical composition in the set of test chemical compositions, a corresponding test vector, thereby obtaining a plurality of test vectors, wherein each corresponding test vector comprises a plurality of test elements, each element in the plurality of test elements comprising a test distribution metric of a respective dimension reduction component in a plurality of dimension reduction components computed using a measurement of a plurality of features across a plurality of test instances of the cell upon exposure of the plurality of test instances of the cell to the respective test chemical composition wherein the plurality of dimension reduction components of the test distribution metric is less than the plurality of features;
• (B) computing a composite test metric, using the plurality of test vectors, wherein the composite test metric quantifies the on target effect of the set of test chemical compositions against the first gene, and wherein the composite test metric is computed by a first process comprising:
• (a) for each respective test vector in the plurality of test vectors, computing a test similarly metric between (i) the respective test vector and (ii) a measure of central tendency of the plurality of test vectors with the respective test vector removed from the plurality of test vectors, thereby obtaining a plurality of test similarity metrics for the plurality of test vectors, each test similarity metric in the plurality of test similarity metrics uniquely corresponding to a chemical composition in the first set of chemical compositions, and
• (b) computing the composite test metric as a measure of central tendency of the plurality of test similarity metrics;
• (C) computing the null distribution, wherein the null distribution comprises a plurality of composite control metrics, each respective composite control metric in the plurality of composite control metrics computed by a second process comprising:
• (a) selecting a respective set of control chemical compositions from a plurality of chemical compositions, wherein each control chemical composition in the respective set of control chemical compositions is against a different gene in the plurality of genes;
• (b) obtaining, for each respective control chemical composition in the respective set of control chemical compositions, a corresponding control vector, thereby obtaining a respective plurality of control vectors, wherein each corresponding control vector comprises a plurality of control elements, each control element in the plurality of control elements comprising a control distribution metric of a respective dimension reduction component in a plurality of dimension reduction components computed using a measurement of the plurality of features across the respective plurality of control instances of the cell upon exposure of the respective plurality of control instances of the cell to the respective control chemical composition, wherein the plurality of dimension reduction components of the control distribution metric is less than the plurality of features;
• (c) computing the respective composite control metric, using the respective plurality of control vectors, wherein the respective control metric quantifies the on target effect of the respective set of control chemical compositions against the corresponding different gene, and wherein the control metric is computed by a third process comprising:
• (1) for each respective control vector, computing a control similarity metric between (i) the respective control vector and (ii) a measure of central tendency of the respective plurality of control vectors with the respective control vector removed from the respective plurality of control vectors, thereby obtaining a plurality of control similarity metrics for the respective plurality of control vectors, each control similarity metric in the plurality of control similarity metrics uniquely corresponding to a chemical composition in the respective set of control chemical compositions, and
• (2) computing the respective composite control metric as a measure of central tendency of the plurality of control similarity metrics;
• and
• (D) comparing the composite test metric to the null distribution, wherein when the composite test metric is differentiated from the null distribution, the set of test chemical compositions is deemed to discriminate the on target effect against the first gene over the null distribution.

A nontransitory computer readable storage medium and one or more computer programs embedded therein for determining whether a set of test chemical compositions discriminates over a null distribution for an on target effect against a first gene of cell, wherein the set of test chemical compositions comprises a plurality of test chemical compositions against the first gene and the cell comprises a plurality of genes including the first gene, the one or more computer programs comprising instructions which, when executed by a computer system, cause the computer system to perform a method comprising:

• (A) obtaining, for each respective test chemical composition in the set of test chemical compositions, a corresponding test vector, thereby obtaining a plurality of test vectors, wherein each corresponding test vector comprises a plurality of test elements, each test element in the plurality of test elements comprising a test distribution metric of a respective dimension reduction component in a plurality of dimension reduction components computed using a measurement of a plurality of features across a plurality of test instances of the cell upon exposure of the plurality of test instances of the cell to the respective test chemical composition, wherein the plurality of dimension reduction components of the test distribution metric is less than the plurality of features;
• (B) computing a composite test metric, using the plurality of test vectors, wherein the composite test metric quantifies the on target effect of the set of test chemical compositions against the first gene, and wherein the composite test metric is computed by a first process comprising:
• (a) for each respective test vector in the plurality of test vectors, computing a test similarly metric between (i) the respective test vector and (ii) a measure of central tendency of the plurality of test vectors with the respective test vector removed from the plurality of test vectors, thereby obtaining a plurality of test similarity metrics for the plurality of test vectors, each test similarity metric in the plurality of test similarity metrics uniquely corresponding to a chemical composition in the first set of chemical compositions, and
• (b) computing the composite test metric as a measure of central tendency of the plurality of test similarity metrics;
• (C) computing the null distribution, wherein the null distribution comprises a plurality of composite control metrics, each respective composite control metric in the plurality of composite control metrics computed by a second process comprising:
• (a) selecting a respective set of control chemical compositions from a plurality of chemical compositions, wherein each control chemical composition in the respective set of control chemical compositions is against a different gene in the plurality of genes;
• (b) obtaining, for each respective control chemical composition in the respective set of control chemical compositions, a corresponding control vector, thereby obtaining a respective plurality of control vectors, wherein each corresponding control vector comprises a plurality of control elements, each control element in the plurality of control elements comprising a control distribution metric of a respective dimension reduction component in a plurality of dimension reduction components computed using a measurement of the plurality of features across a respective plurality of control instances of the cell upon exposure of the respective plurality of control instances of the entity to the respective control chemical composition, wherein the plurality of dimension reduction components of the control distribution metric is less than the plurality of features;
• (c) computing the respective composite control metric, using the respective plurality of control vectors, wherein the respective control metric quantifies the on target effect of the respective set of control chemical compositions against the corresponding different gene, and wherein the control metric is computed by a third process comprising:
• (1) for each respective control vector, computing a control similarity metric between (i) the respective control vector and (ii) a measure of central tendency of the respective plurality of control vectors with the respective control vector removed from the respective plurality of control vectors, thereby obtaining a plurality of control similarity metrics for the respective plurality of control vectors, each control similarity metric in the plurality of control similarity metrics uniquely corresponding to a chemical composition in the respective set of control chemical compositions, and
• (2) computing the respective composite control metric as a measure of central tendency of the plurality of control similarity metrics;
• and
• (D) comparing the composite test metric to the null distribution, wherein when the composite test metric is differentiated from the null distribution, the set of test chemical compositions is deemed to discriminate the on target effect against the first gene over the null distribution.