Classification of samples based on methylation analysis of dna fragments

A method for diagnosing a disease comprising:

• receiving, at one or more processors, sequenced fragment data for a plurality of sequenced fragments derived from a sample from a subject;
• determining, using the one or more processors, a methylation state for each sequenced fragment of the plurality of sequenced fragments based on the sequenced fragment data;
• determining, using the one or more processors, for each sequenced fragment of the plurality of sequenced fragments a probability that the methylation state of a sequenced fragment is significantly different from a distribution of methylation states determined for a plurality of sequenced fragments derived from samples from healthy individuals that map to a same genomic interval;
• determining, using the one or more processors, a disease probability metric for the sample based on the probabilities determined for the plurality of sequenced fragments that the methylation state of a given sequenced fragment is significantly different from the distribution of methylation states determined for the plurality of sequenced fragments derived from samples from healthy individuals;
• and
• outputting, using the one or more processor, a diagnosis that the subject has the disease based on a comparison of the disease probability metric for the sample to a first predetermined threshold.

A method for detecting 3′-end hypomethylation bias, the method comprising:

• receiving, at one or more processors, sequenced fragment data for a plurality of sequenced fragments derived from a sample from a subject;
• determining, using the one or more processors, a methylation status of one or more sites located proximal to a 3′-end of each sequenced fragment of the plurality of sequenced fragments based on the sequenced fragment data, wherein a site at which the methylation status is determined is a methylation site;
• and
• for each sequenced fragment for which an unmethylated status is determined for the one or more methylation sites located proximal to the 3′-end:
• comparing, using the one or more processors, the methylation status of the one or more methylation sites located proximal to the 3′-end of a sequenced fragment under test to that for other sequenced fragments of the plurality that overlap the same one or more methylation sites;
• generating a contingency table that tabulates the results of the comparisons of the methylation status of the one or more methylation sites located proximal to the 3′-end of a sequenced fragment under test to that for the other sequenced fragments of the plurality that overlap the same one or more methylation sites;
• and
• performing a statistical test to determine a probability that an association between two or more factors used to construct the contingency table is significant;
• comparing the determined probability to a second predetermined threshold;
• and
• detecting 3′-end hypomethylation bias in the sequenced fragment under test based on the comparison.