Aligning　protein　sequences　using　a　score　matrix　has　became　a　routine　but　valuable　method　in　modern　biological　research.　However,　alignment　in　＇Twilight　Zone＇　is　still　a　challenging　issue.　New　score　matrices　such　as　Position　Specific　Score　Matrix　and　structural　context-based　matrix　have　been　issued　to　deal　with　it,　whereas　generalized　matrices　are　still　useful　in　pairwise　alignment　and/or　some　steps　of　generating　profiles.　Family-specific　matrices　have　been　proved　effective　in　improving　alignment,　whereas　the　generalization　of　them　is　reduced.　ALPHASUM,　a　score　matrix　based　on　structure　alignment　of　low　identity　all-alpha　proteins　is　constructed.　Performance　comparison　suggests　substitution　pattern　difference　between　all-alpha　proteins　and　all　proteins　and　indicates　that　alignment　of　all-alpha　proteins　in　＇Twilight　Zone＇　can　be　improved　by　using　alpha　class-specific　matrix　and　gap　penalties.　The　implication　between　structure-derived　matrix　and　influence　of　amino　acid　substitution　to　structure　is　also　suggested.　©　2008　IEEE.