Combining genomic and network characteristics for extended capability in predicting synergistic drugs for cancer
Yi Sun, Zhen Sheng, Chao Ma, Kailin Tang, Ruixin Zhu, Zhuanbin Wu, Ruling Shen, Jun Feng, Dingfeng Wu, Danyi Huang, Dandan Huang, Jian Fei
(+2 others)
2015
Nature Communications
Supplementary Figure. 1: Construction of different "training" and "testing" datasets. There are no overlapped drug pairs between each set of "training" and the corresponding "testing" dataset. Figure. 2: Isobolograms for the 17 agent pairs tested by experiment on MCF7 cell line. The horizontal axis and the vertical axis represent the relative dose of the drugs. The area between the two diagonal dotted gray lines in each plot represents the additive effect, while the triangular region under this
more »
... additive effect area indicates the synergy effect. Triangular points represent the relative concentrations of the combinations resulting in 50% inhibition of cancer cells. The two drugs were used alone or in combinations at 4 different concentration ratios: 4:1, 3:2, 2:3, and 1:4. Therefore, there were four Triangular points for each drug pair. Figure. 3: Isobolograms for the 30 randomly picked agent pairs tested by experiment on MCF7 cell line. The horizontal axis and the vertical axis represent the relative dose of the drugs. The area between the two diagonal dotted gray lines in each plot represents the additive effect, while the triangular region under this additive effect area indicates the synergy effect. Triangular points represent the relative concentrations of the combinations resulting in 50% inhibition of cancer cells. The two drugs were used alone or in combinations at 4 different concentration ratios: 4:1, 3:2, 2:3, and 1:4. Therefore, there were four Triangular points for each drug pair. Supplementary Figure. 4: Result of the validation experiment for the randomly picked agent pairs on ER positive breast cancer cell line MCF7. Each drug with the corresponding maximum concentration used for the single drug in the experiment was recorded in Drug1# and Drug2#. Supplementary Supplementary The Chou and Talalay median-effect method was used to calculate the CI and summarized in a heat map. Green indicates synergy (CI < 0.9); yellow indicates additive (0.9 1.1). "4+1", "3+2", "2+3", and "1+4" indicate the two drugs were used in combinations at four different concentration ratios: 4:1, 3:2, 2:3, and 1:4. Supplementary Figure. 5: True positive rate, hit-rate, enrichment-factor of the preliminary ranking model of RACS. 10 times cross validation was performed with 10, 15, or 20 positive samples as training sets respectively. The dark dot is the mean value for true positive rate, hit-rate, and enrichment-factor, while the upper bound and lower bound denote the max and min value of them. Supplementary Figure. 6: True positive rate for dataset with 41 positive samples and 5 features. 10 times cross validation was performed with 10, 20, or 30 positive samples as training sets respectively. The dark dot, the upper bound and lower bound represent the mean value, max and min value for true positive rate. 10 positive samples 15 positive samples 20 positive samples Enrichment_factor Our model One-class SVM Supplementary Figure. 10: Assessment of the contribution of individual features and the combined features to model performance. Values are averaged for 10 times. Supplementary Figure. 11: Performance of RACS with different choice of features. The light blue bars denote the performance of RACS using only one single feature, the dark gold ones denote the performance of RACS when one of the 7 significant feature was removed, while the pink bars represent the performance of RACS when one of the 7 insignificant feature was incorporated. Supplementary Figure. 12: Evaluation of RACS performance by sub-selecting drug targets. a. Distribution of drugs with respect to number of their targets. b. Ranking concordance for agents with 2 target proteins. For each of the 33 drugs with 2 target proteins, 117 drug pairs were obtained when combined with the left 117 drugs. The concordance of predicted ranking lists of the 117 drug pairs before and after removing one of the targets were measured using C-index. c. Ranking concordance for agents with 3 target proteins. For each of the 26 drugs with 3 target proteins, the three single targets together with all three 2-targets-combinations were removed respectively each time. The concordance of predicted ranking lists of the 117 drug pairs before and after removing one/two of the targets were measured using C-index. Supplementary Figure. 16: Result of the validation experiment for the bottom ranked agent pairs on NSCLC A549. Each drug with the corresponding maximum concentration used for the single drug in the experiment was recorded in Drug1# and Drug2#. The Chou and Talalay median-effect method was used to calculate the CI and summarized in a heat map. Green indicates synergy (CI < 0.9); yellow indicates additive (0.9 1.1). "4+1", "3+2", "2+3", and "1+4" indicate the two drugs were used in combinations at four different concentration ratios: 4:1, 3:2, 2:3, and 1:4. Supplementary Figure. 17: Maximum non-lethal concentration (MNLC) and LC50 determination of Sorafenib. a. statistic number of animal situations observed after Sorafenib treatment; b. graphic representation of the data from a. Supplementary Figure. 18: Maximum non-lethal concentration (MNLC) and LC50 determination of Tamoxifen. a. statistic number of animal situations observed after Tamoxifen treatment; b. graphic representation of the data from a.
doi:10.1038/ncomms9481
pmid:26412466
pmcid:PMC4598846
fatcat:szirkmd7tjc4figqlcgaolkox4