The present invention is based on the solving of the crystal structure of adenovirus fiber protein knob domain bound to domain 1 of the coxsackie-adenovirus receptor. One aspect of the present invention relates to a mutant adenovirus which has a genome comprising one or more mutations in sequences which encode the fiber protein knob domain, the viral particle encoded by the genome being characterized by a significantly weakened binding affinity for CARD1 relative to wild-type adenovirus. Such mutations may be in sequences which encode either the AB loop, or the HI loop of the fiber protein knob domain. Specific residues and mutations are described. Another aspect of the present invention is a method for generating a mutant adenovirus which is characterized by a receptor binding affinity or specificity which differs substantially from wild type., from an adenovirus which binds CARD1. In the method, residues of the adenovirus fiber protein knob domain which are predicted to alter D1 binding when mutated, are identified from the crystal structure coordinates of the AD12knob:CAR-D1 complex. A mutation which alters one or more of the identified residues is introduced into the genome of the adenovirus, and whether or not the mutant produced exhibits altered adenovirus-CAR binding properties is determined. Mutants produced by this method include those which under physiological conditions, have significantly weakened binding affinity for CARD1 relative to wild type adenovirus and those which bind a receptor other than CARD1, including an engineered receptor. Introduced mutations may produce an amino acid insertion, deletion or substitution in the encoded viral particle, and may serve to alter the conformation of one or more residues of knob which participate directly in D1 binding. Such residues include residues of the AB loop, the CD loop, the DE loop, the FG loop, the E strand and the F strand. Alternatively, the mutation may be directly introduced in a codon encoding the residue of knob which participates directly in D1 binding. Specific residues in the AB loop, the CD loop, the FG loop, the E strand, the F strand, and the DE loop which participate directly in binding are identified. Another aspect of the present invention is a method for identifying an inhibitor of adenovirus binding to CAR. In the method, a three-dimensional structure derived by X-ray diffraction from a crystal of adenovirus knob trimer bound to CARD1 is provided and then employed to design or select a potential inhibitor. The potential inhibitor is synthesized and then whether or not the potential inhibitor inhibits adenovirus binding to CAR is determined. Preferred crystal structures and space group symmetry is listed. A set of atomic coordinates which define the three dimensional structure is provided. The potential inhibitor may be designed to interact non-covalently with one or more residues of the adenovirus fiber knob protein domain. Alternatively, the potential inhibitor is designed to interact non-covalently with one or more residues of CARD1. Specific residues for covalent and non-covalent interaction are listed. The potential inhibitor may also be designed to interact non-covalently with residues which line a cavity formed during adenovirus knob trimer/CARD1 binding. The potential inhibitor can be designed by identifying chemical entities or fragments capable of associating with the adenovirus knob trimer, and assembling the identified chemical entities or fragments into a single molecule to provide the structure of said potential inhibitor. Such an inhibitor may be designed de novo or from a known inhibitor. Methods of inhibition include competitive inhibition, non-competitive inhibition and uncompetitive inhibition.