The pursuit of novel therapeutic agents in cancer relies on the identification and validation of molecular targets. Hallmarks of cancer include self-sufficiency in growth signals and evasion from apoptosis; genes that regulate these processes may be optimal for therapeutic attack. Here we describe a loss-of-function screen for genes required for the proliferation and survival of cancer cells using an RNA interference library. We used a doxycycline-inducible retroviral vector for the expression of small hairpin RNAs (shRNAs) to construct a library targeting 2,500 human genes. We used retroviral pools from this library to infect cell lines representing two distinct molecular subgroups of diffuse large B-cell lymphoma (DLBCL), termed activated B-cell-like DLBCL and germinal centre B-cell-like DLBCL. Each vector was engineered to contain a unique 60-base-pair 'bar code', allowing the abundance of an individual shRNA vector within a population of transduced cells to be measured using microarrays of the bar-code sequences. We observed that a subset of shRNA vectors was depleted from the transduced cells after three weeks in culture only if shRNA expression was induced. In activated B-cell-like DLBCL cells, but not germinal centre B-cell-like DLBCL cells, shRNAs targeting the NF-kappaB pathway were depleted, in keeping with the essential role of this pathway in the survival of activated B-cell-like DLBCL. This screen uncovered CARD11 as a key upstream signalling component responsible for the constitutive IkappaB kinase activity in activated B-cell-like DLBCL. The methodology that we describe can be used to establish a functional taxonomy of cancer and help reveal new classes of therapeutic targets distinct from known oncogenes.