Human genome project and Protein 3000 project have produced a lot of data of sequences and 3D-structures, respectively.  Now, the simulation of biological systems are being planned on the basis of those data.  However, the paucity of basic knowledge about proteins is an obstacle to the understanding of biological systems.  We have carried out the analyses of physical properties of amino acid sequences for extracting various kinds of signals which are applicable to unknown sequences.  I will talk on three topics about the signals hided in amino acid sequences.  The first topic is the problem of the analog protein pairs. We focused on the similarity of charge distribution between two proteins of analog pairs.  The second is the statistical analysis of secondary structure breakers.  Pro and Gly are well known as secondary structure breakers, and we found another type of breaker that is the clusters of amphiphilic residues.  The third topic is the accurate prediction of membrane proteins, in which signals are the clusters of hydrophobic residues and amphiphilic ones.