1st International and 10th National Iranian Conference on Bioinformatics

Home Accepted Papers

Application of activity cliffs to virtual screening for identify lactate dehydrogenase inhibitors using machine learning approaches

Paper ID : 1339-ICB10

Authors:

Sedigheh Damavandi *¹, Abbasali Emamjomeh², Fereshte Shiri³

¹فارغ التحصیل ارشد بیوانفورماتیک داشگاه زابل

²آزمایشگاه بیوتکنولوژی محاسباتی و بیوانفورماتیک (CBB)، گروه بیوانفورماتیک، دانشکده علوم، دانشگاه زابل.

³گروه شیمی، دانشکده علوم، دانشگاه زابل

Abstract:

Activity Cliffs (ACs) are groups of structurally identical active chemicals with significant potency variances [1–3]. Since ACs are so common in SAR data [2], it's crucial for modern computer-aided drug design and discovery to be able to deal with them effectively [3–9]. ACs, on the other hand, pose considerable challenges for bioactivity-supervised discovery approaches that presume smooth and continuous structure-activity connections [5]. Many ligand-based approaches have high-performance predictions, indicating that the hypothesis is correct. However, there are certain limitations. They are unable to explain the activity cliff [6], and their results for new compounds are suboptimal. The ACs verified to yield pharmacophore and machine learning models that were equivalent to known ligand- and structure-based pharmacophores in terms of accuracy [4]. We applied the ACs to investigate the protein Lactate dehydrogenase which has a high-resolution crystallographic structure and a number of known inhibitors. The number of ACs was determined in each protein's inhibitor population. The missing edges (e.g. unknown interactions) are predicted using various kinds of Machine Learning models. Then, we have searched a large number of machine learners (MLs) to see if we could link protein features to the existence or absence of ACs in the ligand population. Therefore, by identifying ACs that had not been previously considered, the presence of different atomic shares and the differential effects of power associated with ACs formation are determined, which indicated the occurrence of unknown interactions. Finally, using the information deriving from the activity cliff analysis to suggest how virtual screening protocols might be improved to favor the early identification of potent and selective lactate dehydrogenase inhibitors in molecular databases.

Keywords:

Key words: Activity Cliffs; Machine Learning; Lactate dehydrogenase; virtual screening.

Status : Paper Accepted (Poster Presentation)