Frequently asked questions about COXEN.

  1. What does COXEN do?
  2. How does COXEN work?
  3. How can I use COXEN?
  4. What drugs does COXEN provide predictions for?
  5. What cancer types does COXEN provide predictions for?
  6. I have HG-U95A or HG-U133 Plus 2.0 chips? Can I use COXEN?
  7. I have cDNA data. Can I use COXEN?
  8. I have animal data. Can I use COXEN?
  9. Can I run COXEN on Cell Lines?
  10. Can I download your program and run it myself?
  11. Are my CEL files secure?
  12. Are you serious about using this to predict chemotherapy for humans?
  13. How can the COXEN prediction results be validated?
  14. The results you provided were absolutely (in-)correct!
  15. How can we use COXEN to predict appropriate combinations of chemotherapeutic agents?
  16. I am having problems running COXEN. Where can I get help?

  1. What does COXEN do?
    In a nutshell, COXEN provides the prediction probabilities on which chemotherapeutic agents should be effective for the treatment of individual cancerous tumors, based on gene expression profiles of the tumor samples. We provide a prediction value for each chip, which, we hope, will eventually bring individualized chemotherapy one step forward. The current COXEN prediction on most compounds, however, is a suggestive guideline ONLY for testing on experimental systems such as cancer cell lines and animal models.

  2. How does COXEN work?
    See our     theory page for more information.

  3. How can I use COXEN?
    See our     instructions page for an illustrated guide to using the COXEN service.

  4. What drugs does COXEN provide predictions for?
    We can generate predictions for the compounds listed below with generic names on the left and trade names or synonyms in parentheses on the right. This ability is based on the robust availability of NCI-60 panel dose response information on these compounds. Many other interesting compounds (i.e. Imatinib Mesylate / Gleevec) do not have such information available to us and thus cannot be predicted. However, we have the ability to use and provide predictions from any cell line panel, thus, if investigators have dose response data for their compound of interest on their own cell panel, we would be glad to work with them and provide COXEN predictions for their agents.
    Actinomycin(Dactinomycin,Cosmegen)
    Arsenic Trioxide(Trisenox)
    Asparaginase(Elspar)
    Bleomycin(Blenoxane)
    Bortezomib(Velcade)
    Busulfan(Busulfex/Myleran)
    Carboplatin(Paraplatin)
    Carmustine(BiCNU,BCNU,Gliadel)
    Chlorambucil(Leukeran)
    Cisplatin(Platinol)
    Clofarabine(Clolar)
    Cytarabine(Cytosar-U)
    Dacarbazine(DTIC-Dome)
    Daunorubicin(Cerubidine)
    Docetaxel(Taxotere)
    Doxorubicin(Adriamycin)
    Etoposide(Vepesid)
    Floxuridine(FUDR)
    Fludarabine(Fludara)
    Fluorouracil(Efudex)
    Gemcitabine(Gemzar)
    Hydroxyurea(Hydrea/Droxia)
    Idarubicin(Idamycin)
    Ifosfamide(IFEX)
    Irinotecan(Camptosar)
    Lomustine(CEENU)
    Mechlorethamine(Mustargen)
    Melphalan(Alkeran)
    Mercaptopurine(Purinethol)
    Methotrexate(Trexall)
    Mithramycin(Plicamycin/Mithracin)
    Mitomycin(Mutamycin)
    Mitoxantrone(Novantrone)
    Paclitaxel(Taxol)
    Procarbazine(Matulane)
    Phosphoramide Mustard (Cyclophosphamide)(Cytoxan,Neosar)
    Tamoxifen(Nolvadex)
    Teniposide(Vumon)
    Thioguanine
    Thiotepa(Thioplex)
    Topotecan(Hycamtin)
    Toremifine(Fareston)
    Tretinoin(Vesanoid)
    Triethylenemelamine
    Uracil Mustard
    Vinblastine(Velban)
    Vincristine(Oncovin)

  5. What cancer types does COXEN provide predictions for?
    Currently, breast and bladder cancer, but we plan on expandng this shortly. As this service depends on the public availablity of microarray data sets for different cancer types, we strongly encourage users to make their cancer-related microarray datasets publically available by contributing them to     GEO, or to other public databases.

    In the meantime, if you would like to analyze microarray data from a different type of cancer, you may simply define them as breast or bladder cancer. The prediction results for some compounds may hold for certain types of cancer since they may share common chemosensitivity with bladder and breast cancer. However, we warn that the fidelity of such COXEN predictions can be significantly lower without the correct designation of cancer type.

  6. I have HG-U95A or HG-U133 Plus 2.0 chips. Can I use COXEN?
    We hope to enable support for these types of GeneChips shortly.

  7. I have cDNA data. Can I use COXEN?
    We're working on support for these types of chips.

  8. I have animal data. Can I use COXEN?
    Not yet.

  9. I have U133A data for Cell Lines. Can I use that?
    Sure!

  10. Can I download your program and run it myself?
    Running COXEN effectively requires you to download and process several large data sets and additionally the computational requirements are extensive. Thus currently, it is most feasible for us to run it on our servers than to package it for release. In addition, we are always working to improve the code and this way you can be sure you are running the latest version.

  11. Are my CEL files private?
    We are taking reasonable precautions to ensure the security of your files from other users. Unless specifically required for testing purposes we will not look at your files. You may delete your files from our server at any time using the     dataset management page. We will have a copy in our backups for a short period of time, but will not have a permanent copy.

  12. Are you serious about using this to predict chemotherapy for humans?
    No, by any means! All of the COXEN predictions are currently experimental in nature and require multiple, rigorous validations prior to any use for treating human patients. We have developed and support this site only to help certain cancer investigations on experimental systems and to obtain more validation on these predictions at the moment. We will post updates on our continuing development and progress as they occur.

  13. How can the coxen prediction results be validated?
    The COXEN-predicted probabilities of chemosensitivity or chemotherepeutic response are somewhat relative values subject to each compound and each target cancer patient population subtype. It is thus difficult to define an absolute cutoff value discriminating between responder and non-responder patients, or sensitive and resistant cancer cells. Such cutoff values can be defined once those factors are fixed and the prediction results are further validated in a clinical setting. Therefore, we currently recommend that users validate COXEN prediction results by comparing the distributions of predicted probabilities between responder and non-responder groups, e.g. with a two-sample t-test. Again, we hope that users provide feedback on such independent validation tests in order to improve COXEN prediction accuracy.

  14. The results you provided were absolutely (in-)correct!
    Either way, we would love for you to let us know about it so that we can improve the COXEN implementation. Email us at     results@coxen.org.

  15. How can we use COXEN to predict appropriate combinations of chemotherapeutic agents?
    Prediction of chemotherapeutic agents is challenging and predicting synergistic or antagonistic effects, not to mention toxicities can be even more so. At the moment, we suggest that COXEN users assume the activities of different compounds independent each other for testing such combination effects on their experimental systems. We are actively working on developing COXEN for drug combinations and expect to be reporting on this work in 2008.

  16. I am having problems running COXEN. Where can I get help?
    Contact     support@coxen.org for more information.