|
|
Help Conquer Cancer April 2013 Update
Thanks to overwhelming support by World Community Grid members, and their contributions of CPU and GPU cycles, we are nearing the end of the grid-computation phase of the Help Conquer Cancer project. April 4, 2013, marked the completion of 300 million images.
感谢WCG成员的大力支持和他们捐赠的CPU&GPU运算HZ（译注：不知道这么翻译对不对），HCC项目的分布式计算阶段即将结束。2013年4月4日，（我们）注意到完成了3亿张图片的分析。
Thank you all for continued support. It is exciting to finalize this long project and start getting results we have been waiting for. At the same time – we are embarking on another, even more exciting project.
感谢大家的持续支持。完成这个巨大的项目并获取我们一直等待的结果令人兴奋。同时，我们正着手另一个更加激动人心的项目。
New Binary Classifiers
We recently used Help Conquer Cancer results to train a new set of binary classifiers to complement our existing suite of Random Forest classifiers. Unlike the 11-way classifiers in the last update (one model distinguishes between 11 experimental outcomes), each of these classifiers detects a different event in protein crystallization trials: is-clear, has-phaseseparation, has-precipitation, has-skin-effect, has-crystal, and junk. (The junk detector is used to label experiments where an obvious experimental error has occurred, such as a missing droplet, or contamination by dust.) Classification results for each classifier are shown below:
我们最近使用HCC项目的结果（总结）出一套新的二元分类法并完成(译注：学会计的不懂binary classifier是什么，机器上翻译的是二元分类，虽然我感觉是二进制计算/表示的分类法)令人激动的随即森林分类套件（译注：这又是什么？！）。不像上一次升级使用的11路分类器（一个模型会产生11种不同的结果），每一个分类器会探测（模拟）蛋白质结晶（过程）中的不同事件：清晰、分相、沉淀、集肤效应、结晶和垃圾。（垃圾探测器同来标注根据经验已知的错误，比如缺失的粒子或者灰尘污染）。分类结果如下所示：
Of course, the most interesting classifier in the group is the crystal detector. It detects crystals with 93% precision and 53% recall,meaning that in our test set, 93% of the images chosen by he crystal-detector contained a true crystal, but that only 53% of crystals overall were detected. Classifiers from earlier updates and publications showed lower precision but higher recall (80%). A plot of the ROC curve of the crystal classifier is shown below.
当然，这组分类其中最有趣的是晶体分类器，有着93%的准确率和53%的回放（译注：recall专业名词不懂啊），这意味着在我们的测试套件中，被晶体探测器选中的图像有93%包含一个真正的晶体，但是只有53%的整体晶体能被探测到。之前发布并使用的探测器有较低的准确率和更高的回放（80%）。以下是一个晶体分类器受试者工作特征曲线(ROC curve)：
Towards Project Completion and Future Work
Processing of all work units is nearly complete, with just a few weeks remaining. Much work remains to be done in our lab and in our compute cluster, compiling, analyzing, and mining the results, generating new models, and propagating the advances back into the protein crystallization pipeline. Work will carry on. Meanwhile, the end of Help Conquer Cancer will not be the end of our involvement with the World Community Grid. We are busy developing a new project that will search for prognosticand predictive signatures (sets of genes, proteins, microRNAs, etc.) that help predict patient survival and response to treatment. The project will target multiple cancers. The title, launch date, and further details will be announced at a later date.Some of the relevant related papers:
工作包的计算只要几个星期即可结束。剩下的部分学要在我们的实验室跟计算机集群中完成，汇编、分析、挖掘结果、生成新的模型，并且把得到的最新进展放到蛋白质结晶化管线中。工作仍将继续，在此期间，HCC的结束并不意味着我们跟WCG没有关系了。我们正忙于开发用于研究预测特性（成套的基因、蛋白质、微型RNA等）的新项目并用于帮助预测病人的存活率、改进治疗方法。这个项目的目标是多种癌症。项目名称、发布时间和更多细节会在稍晚的时间公布，以下是相关论文：
Cirilo, P. D. R., Marchi, F. A., Filho, M. C. B., Rocha, R. M., Domingues, M. A. C., Jurisica, I., Pontes, A., Rogatto, S. R. An integrative genomic and transcriptomic analysis rReveals potential targets associated with cell proliferation in uterine leiomyomas, PLoS One, 8(3):e57901, 2013.
Vafaee, F., Rosu, D., Broackes-Carter, F. and Jurisica, I. Novel semantic similarity measure improves an integrative approach to predicting gene functional associations, BMC Sys Biol,7(1):22, 2013.
Goswami, R. S., Atenafu, E. G., Xuan, Y., Waldron, L., Pintor dos Reis, P., Sun, T., Datti, A., Xu, W., Kuruvilla, J., Good, D. J., Lai, R., Church, A. J., Lam, W., Baetz, T., LeBrun, D. P., Sehn, L. H., Farinha, P., Jurisica, I., Bailey, D. J., Gascoyne, R. D., Crump, M., and KamelReid, S. A microRNA signature obtained from the comparison of aggressive to indolent nonHodgkin lymphomas can be usedin the prognosis of mantle cell lymphoma, J Clin Oncol, 2013. In press /2012/453050
Berger T, Ueda T, Arpaia E, Chio II, Shirdel EA, Jurisica I, Hamada K, You-Ten A, Haight J, Wakeham A, Cheung CC, Mak TW. Flotillin-2 deficiency leads to reduced lung metastases in a mouse breast cancer model. Oncogene, doi: 10.1038/onc.2012.499. In press.
Pastrello, C., Otasek, D., Fortney, K., Agapito, G., Cannataro, M., Shirdel, E., Jurisica, I. Visual data mining of biological networks: one size does not fit all, PLoS Comp Biol, 9(1): e1002833. doi:10.1371/journal.pcbi.1002833, 2013.
Fortney, K., Xie, W., Kotlyar, M., Griesman, J., Kotseruba, J., Jurisica, I. NetwoRx: Connecting drugs to networks and phenotypes in S. Cerevisiae, Nucl Acids Res, 41(D1): D720-7, 2013.
Starmans MH, Pintilie M, John T, Der SD, Shepherd FA, Jurisica I, Lambin P, Tsao MS, Boutros PC. Exploiting the noise: improving biomarkers with ensembles of data analysis methodologies, Genome Med, 4(11):84, 2012.
Hammerman, P. S., et al., The Cancer Genome Atlas Research Network. Comprehensive genomic characterization of squamous cell lung cancers. Nature. 489, 519–525, 2012.
McKee, C.M., Xu, D., Cao, Y., Kabraji, S., Allen, D., Kearsmans, V., Beech, J., Smart, S., Hamdy, F., Ishkanian, A., Sykes, J., Pintile, M., Milosevic, M., Kwast, T. van der, Zafarana, G., Ramnarine, R., Jurisica, I., Mallof, C., Lam, W., Bristow, R.G., Muschel, R.J. Protease Nexin 1 modulates prostate adenocarcinoma by regulating the Hedgehog pathway. J Clin Invest, 122(11):4025-36, 2012.
Kotlyar, M., Fortney, F. and Jurisica, I. Network-based characterization of drug-regulated genes, drug targets, and toxicity. Methods, 57(4): 477-485, 2012.Heifets, A., Jurisica, I. Construction of new medicines via game proof search. 26th American Association for Artificial Intelligence Conference on Artificial Intelligence (AAAI-12), AAAI Press, Menlo Park, 1564-1570, 2012.
Wong, S., Cercone,N., Jurisica, I. Characterizing healthy and disease states by systematically comparing differential correlation networks in lung. Advances in Health Informatics Conference, 2012
|
评分
-
查看全部评分
|
发表于 2013-4-12 09:51:41
发表于 2013-4-12 09:54:34
