Oberdieck and Pandharipande conjectured [9] that the partition function of the Gromov-Witten/Donaldson-Thomas invariants of S × E, the product of a K3 surface and an elliptic curve, is given by minus the reciprocal of the Igusa cusp form of weight 10. For a fixed primitive curve class in S of square 2h − 2, their conjecture predicts that the corresponding partition functions are given by meromorphic Jacobi forms of weight −10 and index h − 1. We calculate the Donaldson-Thomas partition function for primitive classes of square -2 and of square 0, proving strong evidence for their conjecture.Our computation uses reduced Donaldson-Thomas invariants which are defined as the (Behrend function weighted) Euler characteristics of the quotient of the Hilbert scheme of curves in S × E by the action of E. Our technique is a mixture of motivic and toric methods (developed with Kool in [5]) which allows us to express the partition functions in terms of the topological vertex and subsequently in terms of Jacobi forms. We compute both versions of the invariants: unweighted and Behrend function weighted Euler characteristics. Our Behrend function weighted computation requires us to assume Conjecture 18 in [5].Date: November 22, 2017. 1 Since this paper was originally written in 2015, Oberdieck and Pixton have recently given a complete proof of this conjecture using methods (different from ours) from both Donaldson-Thomas theory and Gromov-Witten theory [11]. 1 2 JIM BRYANOur general computational strategy is the following. Donaldson-Thomas invariants are given by weighted Euler characteristics of Hilbert schemes. We stratify the Hilbert scheme using the geometric support of the curves and we compute Euler characteristics of strata separately. Many of the strata acquire actions of E or C * (that were not present globally) and we restrict to the fixed point loci. We are able to further stratify the fixed point loci and those strata sometimes acquire further actions. Iterating this strategy, we reduce the computation to subschemes which are formally locally given by monomial ideals. These are counted using the topological vertex. New identities for the topological vertex lead to closed formulas. To incorporate the Behrend function into this strategy, we must assume the conjecture formulated in [5, Conj. 18]. This is a general conjecture regarding the behavior of the Behrend function at subschemes given locally by monomial ideals.Acknowledgements. I'd like to thank George Oberdieck, Rahul Pandharipande, and Yin Qizheng for invaluable discussions. I've also benefited with discussions with Tom Coates, Sheldon Katz, Martijn Kool, Davesh Maulik, Tony Pantev, Balazs Szendroi, Andras Szenes, and Richard Thomas. The computational technique employed in this paper was developed in collaboration with Martijn Kool whom I owe a debt of gratitude. I would also like to thank the Institute for Mathematical Research (FIM) at ETH for hosting my visit to Zürich, and for Matematisk Institutt, UiO and Jan Christophersen for organizing the 2017 Abel S...