NMR offers the possibility of accurate secondary structure for proteins that would be too large for structure determination. In the absence of an X-ray crystal structure, this information should be useful as an adjunct to protein fold recognition methods based on low resolution force fields. The value of this information has been tested by adding varying amounts of artificial secondary structure data and threading a sequence through a library of candidate folds. Using a literature test set, the threading method alone has only a one-third chance of producing a correct answer among the top ten guesses. With realistic secondary structure information, one can expect a 60-80% chance of finding a homologous structure. The method has then been applied to examples with published estimates of secondary structure. This implementation is completely independent of sequence homology, and sequences are optimally aligned to candidate structures with gaps and insertions allowed. Unlike work using predicted secondary structure, we test the effect of differing amounts of relatively reliable data.Keywords: chemical shift index; fold recognition; protein folding; protein structure prediction; protein threading; remote homology detection; secondary structureThere is no shortage of methods for predicting a protein's structure based only on its sequence~Böhm, 1996; Westhead & Thornton, 1998!. Unfortunately, unless the sequence has significant sequence homology to something of known structure, one could not regard any of the methods as reliable~Lesk, 1997; Levitt, 1997; MarchlerBauer et al., 1997!. At the same time, they may be the only means of predicting structure in the absence of experimental data. A different problem arises for a protein sequence when a limited amount of experimental information is available. A typical case might come from a protein that yields a barely useful NMR spectrum. In this situation, one would like to use the available data, even if it is not suitable for conventional high resolution structure calculations. This has led to a series of approaches that have their roots in structure prediction, but attempt to incorporate very sparse experimental data such as a few intramolecular distance estimates~Smith- Brown et al., 1993;Aszódi et al., 1995;Lund et al., 1996; Skolnick et al., 1997!. Typically, these methods produce low resolution structures and operate with the caveat that answers may sometimes be quite wrong.Taking this theme further, NMR data may provide still more low resolution data. Even if a protein's structure will never be solved, its proton and heteronuclear NMR assignments may be largely determined. The relationship between chemical shift and structure has long been recognized~Pardi et al., 1983; Spera & Bax, 1991!, but it can be better quantified. Given a fairly complete set of proton and heteronuclear chemical shifts, one can expect secondary structure assignments to be more than 92% accurate~Wishart et al., 199192% accurate~Wishart et al., , 1992Wishart & Sykes, 1994a!. The aim of ...