Noninvasive discrimination of rejection in cardiac allograft recipients using gene expression profiling.

Rejection diagnosis by endomyocardial biopsy (EMB) is invasive, expensive and variable. We investigated gene expression profiling of peripheral blood mononuclear cells (PBMC) to discriminate ISHLT grade 0 rejection (quiescence) from moderate/severe rejection (ISHLT > or = 3A). Patients were followed prospectively with blood sampling at post-transplant visits. Biopsies were graded by ISHLT criteria locally and by three independent pathologists blinded to clinical data. Known alloimmune pathways and leukocyte microarrays identified 252 candidate genes for which real-time PCR assays were developed. An 11 gene real-time PCR test was derived from a training set (n = 145 samples, 107 patients) using linear discriminant analysis (LDA), converted into a score (0-40), and validated prospectively in an independent set (n = 63 samples, 63 patients). The test distinguished biopsy-defined moderate/severe rejection from quiescence (p = 0.0018) in the validation set, and had agreement of 84% (95% CI 66% C94%) with grade ISHLT > or = 3A rejection. Patients >1 year post-transplant with scores below 30 (approximately 68% of the study population) are very unlikely to have grade > or = 3A rejection (NPV = 99.6%). Gene expression testing can detect absence of moderate/severe rejection, thus avoiding biopsy in certain clinical settings. Additional clinical experience is needed to establish the role of molecular testing for clinical event prediction and immunosuppression management.

Pharmaceutical target discovery using Guilt-by-Association: schizophrenia and Parkinson's disease genes.

We wish to identify genes associated with disease. To do so, we look for novel genes whose expression patterns mimic those of known disease-associated genes, a method we call Guilt-by-Association (GBA). GBA uses a combinatoric measure of association that provides superior results to those from correlation measures used in previous expression analyses. Using GBA, we have examined the expression of 40,000 human genes in 522 cDNA libraries, and have identified several hundred genes associated with known cancer, inflammation, steroid-synthesis, insulin-synthesis, neurotransmitter processing, matrix remodeling and other disease genes. The majority of the genes thus discovered show no significant sequence similarity to known genes, and thus could not have been identified by homology searches. We present here an example of the discovery of five genes associated with schizophrenia and Parkinson's disease. Of the 40,000 most-abundant human genes, these five genes are the most closely linked to the known disease genes, and thus are prime targets for pharmaceutical intervention.

Measuring the strength of side-chain hydrogen bonds in peptide helices: the Gln.Asp (i, i + 4) interaction.

Whether hydrogen bonds between side chains are energetically significant in proteins and peptides has been controversial. A method is given here for measuring these interactions in peptide helices by comparing the helix contents of peptides with 1, 2, or 3 interactions. Results are given for the glutamine--aspartate (i, i + 4) hydrogen-bond interaction. The Gibbs energy of the interaction is -1.0 kcal/mol when aspartate is charged and -0.4(4) kcal/mol when it is protonated. Magnetic resonance experiments show that the aspartate carboxylate group interacts specifically with the trans amide proton (HE) of glutamine. The interaction is observed only when the glutamine residue is N-terminal to the aspartate and when the spacing is (i, i + 4). The same stereochemistry is found in protein structures, where the (i, i + 4) glutamine-aspartate interaction occurs much more frequently than other possible arrangements.

Discovering structural correlations in alpha-helices.

We have developed a new representation for structural and functional motifs in protein sequences based on correlations between pairs of amino acids and applied it to alpha-helical and beta-sheet sequences. Existing probabilistic methods for representing and analyzing protein sequences have traditionally assumed conditional independence of evidence. In other words, amino acids are assumed to have no effect on each other. However, analyses of protein structures have repeatedly demonstrated the importance of interactions between amino acids in conferring both structure and function. Using Bayesian networks, we are able to model the relationships between amino acids at distinct positions in a protein sequence in addition to the amino acid distributions at each position. We have also developed an automated program for discovering sequence correlations using standard statistical tests and validation techniques. In this paper, we test this program on sequences from secondary structure motifs, namely alpha-helices and beta-sheets. In each case, the correlations our program discovers correspond well with known physical and chemical interactions between amino acids in structures. Furthermore, we show that, using different chemical alphabets for the amino acids, we discover structural relationships based on the same chemical principle used in constructing the alphabet. This new representation of 3-dimensional features in protein motifs, such as those arising from structural or functional constraints on the sequence, can be used to improve sequence analysis tools including pattern analysis and database search.

Determination of free energies of N-capping in alpha-helices by modification of the Lifson-Roig helix-coil therapy to include N- and C-capping.

We have previously shown that varying the N-terminal amino acid in alpha-helical peptides can cause large variations in helix content (Chakrabartty et al., 1993a). The Lifson-Roig theory for the helix-coil transition predicts, however, that substitutions at the N-terminus in an unacetylated peptide should have no effect on alpha-helix stability. We have therefore modified the theory to include these N-capping effects by assigning a statistical weight (the "n-value") to the amino acid immediately preceding a stretch of helical residues. The n-value measures the N-capping propensity of an amino acid, and like the helix propensity (w-value), it is independent of neighboring residues or positions in sequence. The new theory was used, with the experimental data for these substitutions, to calculate n-values and, hence, free energies for N-capping for the amino acids Gln, Ala, Val, Met, Pro, Ile, Leu, Thr, Gly, Ser, and Asn as well as for the acetyl group, which is commonly used to cap peptides. The free energies vary by approximately 1 kcal mol-1 from Gln (worst) to Asn (best), and the acetyl group is nearly as effective as Asn. N-Capping free energies were also found for Leu, Thr, Gly, Ser, and Asn when the N-terminus is charged at pH 5. The unfavorable effect of protonation of the N-terminus in an alpha-helix was found to be approximately 0.5 kcal mol-1. Our results agree well with a survey of N-capping preferences from protein crystal structures and are compared to results from site-directed mutagenesis of N-caps in proteins.

Discovering side-chain correlation in alpha-helices.

Using a new representation for interactions in protein sequences based on correlations between pairs of amino acids, we have examined alpha-helical segments from known protein structures for important interactions. Traditional techniques for representing protein sequences usually make an explicit assumption of conditional independence of residues in the sequences. Protein structure analyses, however, have repeatedly demonstrated the importance of amino acid interactions for structural stability. We have developed an automated program for discovering sequence correlations in sets of aligned protein sequences using standard statistical tests and for representing them with Bayesian networks. In this paper, we demonstrate the power of our discovery program and representation by analyzing pairs of residues from alpha-helices. The sequence correlations we find represent physical and chemical interactions among amino-acid side chains in helical structures. Furthermore, these local interactions are likely to be important for stabilizing and packing alpha-helices. Lastly, we have also detect correlations in side-chain comformations that indicate important structural interactions but which don't appear as sequence correlations.

Detection of correlations in tRNA sequences with structural implications.

Using an flexible representation of biological sequences, we have performed a comparative analysis of 1208 known tRNA sequences. We believe we our technique is a more sensitive method for detecting structural and functional relationships in sets of aligned sequences because we use a flexible representation (for sequences), as well as a general statistical method that can detect a wide range of relationships between positions in a sequence. Our method utilizes functional classifications of the sequence building-blocks (nucleotide bases and amino acids) based on physical or chemical properties. This flexibility in sequence representation improves the significance of finding sequence relationships mediated by the defining property. For example, using a purine/pyrimidine classification, we can detect base-stacking interactions in sets of nucleotide sequences that form base-paired helices. We use several statistical measures, including chi 2-tests, Monte Carlo simulations and an information measure to detect significant correlations in sequences. In this paper we illustrate our method by analyzing a set of tRNA sequences and showing that the correlations our program discovers, in each case, correspond to the known base-pairing and higher order interactions observed in tRNA crystal structures. Furthermore, we show that novel and interesting features of tRNAs are detected when sequence correlations with the charged amino acid (and anticodon) are evaluated. This technique is a powerful method for predicting the structure of RNAs and for analyzing specific functional characteristics.