This article describes revisions to four of the eight constructs of the Duncan molecular genetics learning progression [Duncan, Rogat, & Yarden, (2009)]. As learning progressions remain hypothetical models until validated by multiple rounds of empirical studies, these revisions are an important step toward validating the progression. Our revisions are based on empirical data obtained from tenth grade students in three classroom contexts (n = 121); although our study was done with students at the upper bounds of the progression, students held naive ideas prior to instruction which allowed us to track their ideas through all the levels of each construct during the course of one academic year. We revised the four constructs that center around the molecular model of genetics using students’ pre/post assessments and interviews. We found that tenth grade students do hold ideas consistent with the hypothesized levels in the progression as well as several intermediate ideas not included. Our revisions include adding student ideas that are important conceptual stepping stones in each construct as well as other modifications such as splitting and combining levels, moving ideas to other constructs, changing the conceptual progression of a construct and splitting a construct. Along with the revisions, we identified challenges in each construct. Even after instruction, students had difficulties understanding that genes code for proteins, that proteins connect genes and traits, and how differential gene expression leads to different repertoires of proteins inside of specialized cells. Our findings indicate that classroom instruction should focus more on proteins: how they are created, what their functions are, and how cells express different proteins. © 2015 Wiley Periodicals, Inc. J Res Sci Teach 53: 1385–1418, 2016